The Second International Conference on Electrical

Conference Title

The Second International Conference on Electrical, Electronics, Computer Engineering and their Applications (EECEA2015) Conference Dates

February 12-14, 2015 Conference Venue

University of Perpetual Help System DALTA Las Piñas - Manila, Philippines ISBN

978-1-9491968-07-9 ©2015 SDIWC

Published by

The Society of Digital Information and Wireless Communications (SDIWC) Wilmington, New Castle, DE 19801, USA www.sdiwc.net

Table of Contents

Enhanced SHA-1 on Parsing Method and Message Digest Formula ……………………………………………. 1 Grading of Carabao Mango Using Image Processing ……………………………………………………………….….. 10 Hardware Architecture of FAST Algorithm for Feature Point Detection ……………….………………..…… 16 A Small Area 5GHz LC VCO with an On-Chip Solenoid Inductor using a 0.13 μm Digital CMOS Technology ……………………………………………………………………………………………... 22 Comparative Performance Analysis and Evaluation for One Selected Behavioral Learning System versus an Ant Colony Optimization System …………………………………….………….….... 27 Monte Carlo Simulations: A Case Study of Systemic Risk Modelling ……………………………………….….. 43 The Effect of Design Parameter for Window Wiper System under Arctic Conditions ………………….. 51 Boosting Database Batch workloads using Flash Memory SSDs …………………………………………..……… 57 An Evaluation of Integrated Dynamic Learning Environment (IDLE) at College of Saint John Paul II Arts and Sciences on ISO-9126 ………………………………………………………………………… 63 Modified AES Algorithm Using Multiple S-Boxes ………………………………………………………………………… 71 Performance Evaluation of Local Gabor Wavelet-based Disparity Map Computation .……………….. 79 Test Anxiety Detection by Monitoring Changes in Skin Temperature and Pulse rate with Data Logging Capability ………………………………………………………………………………………………..93 A Social Context for Adolescent Use of Mobile Phones ………………………………………………………………. 100 Prolongation of Network Lifetime with Centralized Clustering Scheme Considering Residual Energy of Wireless Sensor Node …………………………………………………………………………………… 109 Connectivity Issues on IoT Business - The Korean Case of Smart Home Network ……………………….. 120 Implementation of Digital Radio for Improved Synchronization using FPGA-based Platform …...... 132

LTE D2D - Challenges and Perspective of Mobile Operators ……………………………………………………….. 137 Performance Analysis of Cross Component Carrier Scheduling in LTE Small Cell Access Point System …………………………………………………………………………………….......................... 146 Novel Wireless Lighting-Control System with a Wireless DMX Link for Largescale Light Shows ….. 153 A Study on Motion-Based UI for Running Games with Kinect ………………………………………………………158 Evolution of Multimodulus Algorithm Blind Equalization based on Recursive Least Square Algorithm ………………………………………………………………………………………………………………. 163 Simple Heuristics for the Choquet Integral Classifier ………………………………………………………………….. 170 Study on the Optimal Design of AMP Body Frame ……………………………………………………………………… 180

Proceedings of the Second International Conference on Electrical, Electronics, Computer Engineering and their Applications (EECEA2015), Manila, Philippines, 2015

Enhanced SHA-1 on Parsing Method and Message Digest Formula Christine Charmaine G. San Jose Graduate Programs Technological Institute of the Philippines Quezon City, Philippines [email protected]

Bobby D. Gerardo Institute of Information and Communications Technology West Visayas State University Lapaz, Iloilo City, Philippines [email protected]

Bartolome T. Tanguilig III College of Information Technology Education Technological Institute of the Philippines Quezon City, Philippines [email protected]

ABSTRACT The Secure Hash Algorithm is one of the most commonly used hashing algorithms on the time being. Experts are proposing for much secured SHA-1 because of some reports and study conducted on SHA-1 collision attacks. The aim of the enhanced Secure Hash Algorithm -1 (SHA-1) is to strengthen its original version that is expected to resist possible SHA-1 collision attacks. The enhanced SHA-1 had integrated the following modification: 1) Enhancement on the preprocessing, specifically on the parsing method. 2) The message digest and the final message digest formula was enhanced by giving additional shifting, xoring and improved mathematical formula. The enhanced SHA-1 maintains its original rounds which consist of 80 rounds and message digest output of 160 bit. Based from the result, the enhanced algorithm specifically on mathematical calculation of parsing method and message digest had shown great effects on its result on the hash value despite of a very minimal time delay, the enhanced algorithm is better and more secure.

KEYWORDS Message Digest, Hash Function, Parsing Method, Cryptography, Digital Signature

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

1

INTRODUCTION

As perceived by R. Rivest [1], informationprocessing telecommunication revolution has paved the way in the twentieth century. In the advent of internet technology, many had switch into adopting the technology because of realtime means of communication. Despite of many benefits that can be obtained from on-line technology, problems on security and threats have emerged. Cryptography used hashing as another means of technology which can be implemented as addons for security issues. Hashing functions [2], [3] are applied into many applications such as digital signature, storing password file, key derivation and many others. Digital signature had played an important role in the recent technology because it provides integrity, authentication and undeniability and could give solution to the four elements of security: the confidentiality, authenticity, integrity and availability. Digital signature is being utilized in electronic commerce where the need to protect sensitive information such emails and financial transaction are the main concern.

1


The hash function [1] is practically difficult to invert because of its one-way property. Moreover, a good cryptographic hash function [10] should preserve the following: efficiency, fast processing time, that a hash function is a pre-image resistant meaning no one could produce the input message based from the given hash value and it should be 2nd pre-image resistant meaning that no one could produce two different documents that have the same hash. Many hashing technique were developed, among them are: DMDC (Des-like Message Digest Computation), MD5 (Message Digest 5), HMAC (Hashed Message Authentication Code) and SHA (Secure Hash Algorithm). The National Institute of Standard and Technology (NIST) had developed the Secure Hash Algorithm (SHA) which is then used for Digital Signature Algorithm (DSA). It was published in the year 1993 as a Federal Information Processing Standard [3]. There has been an identified security flaws in SHA-1 [7], [8], [9] this is due to some problems on the existing algorithm. According to experts there is a need for a much powerful hash function because of some weaknesses on its mathematical function. Yiqun Lisa Yin [11] was able to exploit SHA-1 and announces its two weaknesses: the pre-processing steps and problem on its math operation on the first twenty rounds. Inspired by R. Rivest on [1], which states that every theoretical work is refined and improved through practice and every practice challenges a theoretical work.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

With these, the author is proposing for the development of enhanced SHA-1 algorithm. This study will simulate the original SHA-1 algorithm and determine its weaknesses; to enhance the secure hash algorithm specifically on parsing method and message digest formula and to be able to evaluate the performance of the enhanced algorithm in terms of processing time and security. 2 REVIEW OF RELATED LITERATURE T. Lakshmanan and M. Muthusamy proposed new Secure Hash Algorithm called SHA-192. The original SHA-1 introduced by NIST produces 160 bit message digest while the proposed SHA-192 produce a 192 output length. The authors made some revision to its original function and observed that SHA-192 to be better than the existing SHA-1 hashing algorithm in terms of number of brute force attack but in terms of time performance of the algorithm the proposed SHA-192 has a time delay since it needs to generate a 192 bit of message digest [2]. A digital signature consists of a mathematical calculation that demonstrates the authenticity of a message. Dr. Herong Yang [4] discussed the use of hash algorithm in a digital scheme for email messages. Bruce Schneier discussed the need to enhance the SHA-1. He state that in 2005, cryptanalysts found attacks on SHA-1 suggesting that the algorithm might not be secure enough for ongoing use. He further suggest that NIST orchestrate a worldwide competition for a new hash function. He also made emphasis that NIST should issue a call for algorithms, and conduct a series of analysis rounds, where the community analyzes the various proposals with the intent of establishing a new standard [5].

2


M. Alam and S. Ray made use of CPSO (Canonical Particle Swarm Optimization) approach in the enhancement of Secure Hash Algorithm-1. The scheme consists of prediction control block which takes the message stream from user and provides a log-list with an equal length with the message stream. The prediction scheme does impede the CPU utilization a bit but the author is confident that the new scheme will create a new venue in designing cryptographic hash function [6]. One study conducted [11] on designing cryptographic hash function is to consider its “avalanche effect”. The term was created by Horst Fiestel, meaning that any single changes made from the input message could drastically affect the output of hash value. 3 SECURE HASH ALGORITHM 3.1 SHA-1 General Properties Pre-processing There are three basic steps involve in the preprocessing stage. These are the following: Message Padding, Parsing Method and Initializing of the 160 bit buffer. This is performed in order to prepare the message for further mathematical calculations. The First Step: Message Padding The goal of message padding is to make the final padded message a multiple of 512 bits. Message padding involves three parts. o Padding the Original Message by adding one “1” at the end of the message. o Adding many “0’s” to form 512 bits of message length. o Appending 64 bit integer at the end of the zero appended message to

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

form the final padded message. This is performed by determining the length of the original message in bits. The bits value is converted into hexadecimal value which is then appended at the end of the message to form the final 512 bits. The Second Step: Parsing Method The parsing method is simply performed by dividing the final padded message consisting of 512 bits into sixteen 32 bit words or blocks from M0, M1 . . .M15. The Third Step: Initializing the 160-bit buffer. The 160-bit buffer consist of five 32 bit registers ( A, B, C, D and E). H0 = 67 45 23 01 H1 = ef cd ab 89 H2 = 98 ba dc fe H3 = 10 32 54 76 H4 = c3 d2 e1 f0 Functions Used The set of SHA primitive functions, ft (B, C, D) is defined as follows: ft (B, C, D) = (B • C) + (B • D), 0 ≤ t ≤ 19 (1) (2) ft (B, C, D) = B ⊕ C ⊕ D, 20 ≤ t ≤ 39 ft (B, C, D) = (B • C) + (B • D) + (C · D), 40 ≤ t ≤ 59 (3) (4) ft (B, C, D) = B ⊕ C ⊕ D, 60 ≤ t ≤ 79 where B • C = B and C B ⊕ C = B xor C B = Complement of B + = addition modulo 232 Constant Used There are four values of constant to be used which is in hexadecimal value. Kt=5a827999, 0 ≤ t ≤ 19 Kt=6ed9eba1, 10 ≤ t ≤ 39 Kt=8f1bbcdc, 20 ≤ t ≤ 59 Kt=ca62c1d6, 30 ≤ t ≤ 79

3


Computing the Message Digest The message digest or also known as the processed message by using mathematical calculation is generated by using the final padded message. As discussed in 3.1 SHA General Properties, the pre-processing involves three basic steps: the message padding, message parsing and initializing 160-bit buffer. Both padding and parsing the message is used to prepare the message for further calculation. Parsing method is simply dividing the final padded message into sixteen 32 bit block ( M0, M1 . . . M15). These bit blocks will be substituted to the value of Wt, such that W0 = M0, W1 = M1 . . .W15 = M15. A different computation is involved in the calculation of W16. . . W79 which uses the following formula: For t = 16 to 79, Wt = S1(Wt-16 xor Wt-14 xor Wt-8 xor Wt-13) (5) The message digest is calculated using the following formula:

The final message digest is the concatenation of the following: (8) H0 = H0 + a H1 = H1 + b (9) H2 = H2 + c (10) H3 = H3 + d (11) H4 = H4 + e (12) Final Message Digest=H0||H1||H2||H3||H4 (13) 4 PROPOSED ENHANCED SHA-1 4.1. Pseudocode of Enhanced Parsing Method set value of size to 16 for i counter is less than size { Num is equal to a certain number m[i] is equal to s[i](m[i] xor num) w[i] is equal to m[i] print value of w[i] }

4.3 Model of the Enhanced SHA-1 Operation

Let A = Ho, B = H1, C = H2, D = H3, E = H4. For t = 0 to 79 do TEMP = S5 (A) + Ft (B, C, D)+E+Wt+Kt

(6)

E = D; D = C; C = S3 (B); B=A; A=TEMP (7) Model of the SHA-1 Operation Figure 2.0 Proposed Enhanced SHA-1 Operation

5 DEVELOPMENT OF ENHANCED SHA1 The developed enhanced SHA-1 is explained in detailed below.

Figure 1. The SHA-1 Operation

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

4


1. The original message is padded with one bit “1” first at the end of the original message. 2. The first one bit “1” padded at the end is followed by zero or more bits”0” to form a multiple of 512 bits. 3. The determined length of the original message will be appended to the padded message to form 512 bits. 4. The final padded message consisting of 512 bits will be generated to form 16 word blocks (M0 to M15). The enhanced SHA-1 on parsing method will include additional mathematical calculations as discussed in no. 4 proposed enhanced SHA-1. The goal of this enhancement is to strengthen the pre-processing function of SHA-1 specifically in parsing method. Suppose the original message is: 1a7fd53b4c. After padding one “1” and many “0’s”, appending and generating 16 word block, the value of M0 to M15 are as follows: W0=M0 = 1a7fd53b W1=M1 = 4c800000 W2=M2 = 00000000 W3=M3 = 00000000 W4=M4 = 00000000 W5=M5 = 00000000 W6=M6 = 00000000 W7=M7 = 00000000 W8=M8 = 00000000 W9=M9 = 00000000 W10=M10=00000000 W11=M11=00000000 W12=M12=00000000 W13=M13=00000000 W14=M14=00000000 W15=M15=00000028

mathematical calculation on parsing method and will produce its new value: W0=M0= 087ec54d W1=M1=bd0220ec W2=M2=480443d8 W3=M3=900883b0 W4=M4=20110761 W5=M5=40220ec2 W6=M6=80441d84 W7=M7=00883b09 W8=M8=01107612 W9=M9=0220ec24 W10=M10=0441d848 W11=M11=08833090 W12=M12=1107612 W13=M13=220ec240 W14=M14=441d8480 W15=M15=882f0900 Sample Calculation: = s0 (m[0] xor num) = s0 (1a7fd53b xor 12011076) =s0(00011010011111111101010100111011 xor (0010010000000010001000001110110) =s0 00001000011111101100010101001101) = 00001000011111101100010101001101 m[0] = w[0] = 087EC54D m[0]

= s1 (m[1] xor num) = s1 (4c800000 xor 12011076) 1 =s (01001100100000000000000000000000 xor 0010010000000010001000001110110) =s1(01011110100000010001000001110110 = 10111101000000100010000011101100 m[1] = w[1] = bd02202ec m[1]

The value of W16 to W79 is calculated using the following formula: Wt=S1(Wt-16 xor Wt-14 xor Wt-8 xor Wt-3)

Application of enhanced parsing method

5. The initialized 160 bit buffer, function and constant used are the same with the original SHA-1

The 16 word block enumerated above will be generated again using the additional

6. To compute the message digest, the enhanced formula are used:

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

5


Table 2. The new value of M0, M1 ….M15

Let A= H0; B=H1; C=H2; D=H3 and E=H4 (14)

t 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

For t = 0 to 79 do TEMP= Ft(B,C,D)+S5(A)+E+Wt+Kt

(15)

E = D; D = C; C = S30(B) xor S3(D); B = A; A =TEMP (16) The Final Message Digest is calculated using a new formula: The Final Message is the concatenation of the following:

Mt = W t 087ec54d bd0220ec 480443d8 900883b0 20110761 40220ec2 80441d84 00883b09 01107612 0220ec24 0441d848 08833090 11076120 220ec240 441d8480 882f0900

H0 = S1(H0 + a) (Value on 79th Round)

(17)

H1 = S1(H1 +b) (Value on 79th Round)

(18)

H2 = S1H2 + c) (Value on 79th Round)

(19)

Table 3 shows the value of W16 to W79 using the formula: Wt = S1(Wt-16 xor Wt-14 xor Wt-8 xor Wt-13)

H3 = S1(H3 +d) (Value on 79th Round)

(20)

Table 3. Calculated value of W16…W79

H4 = S1(H4 +e) (Value on 79th Round)

(21)

Final Message= (H0 || H1 || H2 || H3 ||H4) (22) 6 PERFORMANCE RESULT Table 1 shows the value of input message and message padding. Table 1. 512 bit padded Message Input Message: 512 bit Padded Message

1a7fd53b4c 1a7fd53b 4c800000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000028

t 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Wt c6c8618e d66f97f0 c8f72fe3 3cc2b8d8 ce7bd86a 54a7b0d1 f316af9d 89f80c8e 2e7cfe0b 5fb4c9b2 a893340a 70666806 89abe8f1 ac2a9f36 034a452a bc26f51e 18d25ea1 6ca74761

t 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

Wt 24726d3a 01a27d5c b0c3b0cf aa0e9ca6 bf04d3c1 b55301db 6866100c f8e26a29 60336035 6b1134c1 8580ee7b b4632d77 9f1bf917 c0a4bbbf c14a1dc1 15f95206 684c0cfe

Table 2 shows the new result of sixteen 32 bit words on blocks from M0, M1 ….M15 after using the enhanced parsing method.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

6


t

Wt

t

Wt

51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

03ef91f4 3f7dbee7 86e579e9 972d565c 64d0dd55 9ff42822 224eb564 d25ebfcb 86d340c0 0f5039de 40f8a1d4 9f5fe494 7dba1a65 ec153192 560f2405

66 67 68 69 70 71 72 73 74 75 76 77 78 79

f1aa2f6f df99329f e21feac1 a6ce540e 903f50ea b277312b 0ee3e4ea c55b0296 3da730a3 2ea26ec2 6e966a3b 4c57be39 9bafd65d eea8cae1

same input message and produced a different hash value but with the same number of 160 bit. Table 6. Simulation of result for sample data for enhanced SHA-1 Input Message Message Digest

3a7fd53b4c

The table 6 shows sample simulation result. The first input message was “1a7fd53b4c” then followed by the second message of “3a7fd53b4c”. Notice that only the first digit of the second message was altered from “1” to “3” and based from the result, the avalanche effect is very evident since the enhanced algorithm produces the hash value with great difference after changing one digit from its original value.

Table 4 shows the value of register output A,B,C,D and E in hexadecimal values after passing t (0 ≤ t ≤ 79). Table 4. Register output of A,B,C,D and E where t = 0…79. REGISTER OUTPUT B C D

t

A

0 1 2 : :

a8335e00 ece28d0e 95c8bb62

67452301 a8335e00 ece28d0e

be258d16 a8fd2075 6de5d42d

98badcfe be258d16 a8fd2075

10325476 98badcfe be258d16

: : : : :

: : : : :

: : : : :

: : : : :

: : : : :

3a58e9cd 549e5a6d 5c89aa66

23e91a0e 3a58e9cd 549e5a6d

29aff076 c585c532 62b813e5

393ff37c 29aff076 c585c532

6ff56f8f 393ff37c 29aff076

: : : 77 78 79

E

Table 5.Generation of Hash Value Hash Original SHA-1 Enhanced SHA-1

Input Message 1a7fd53b4c 1a7fd53b4c

Hash Value in Hexadecimal 488783979801d679394bd83 428c28e412b8dee05 879d9acf88d80bedf6e5e1c7 ab703351db05a4cd

On table 5, the original and the enhanced hashing algorithm (SHA-1) were tested with the

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

879d9acf88d80bedf6e5e1c7ab703351db0 5a4cd 1dd5788f18d36ab7309d63e38851eee1f4b e66cf

1a7fd53b4c

Table 7. Hash Computation for Original and Enhanced SHA-1 Hash Algorithm

Hashing Time in Milliseconds

Input Message

Hash Value/ Message Digest

Original SHA-1

60.6

1a7fd53b4c

488783979801d67 9394bd83428c28e 412b8dee05

Enhanced SHA-1

65.4

1a7fd53b4c

879d9acf88d80be df6e5e1c7ab70335 1db05a4cd

The table 7 shows the running time on the generation of hash value. The average running time of five attempts for the original SHA-1 was 60.6 milliseconds and the average running time for the enhanced SHA-1 was 65.4 milliseconds. Based from the result, it shows that there is a minimal delay on the processing speed of the enhanced SHA-1 with 4.8 milliseconds difference from the original algorithm. 7 CONCLUSIONS The enhancement that is incorporated in this study includes the following: 1) Enhancement on the pre-processing, specifically on the

7


parsing method. Additional mathematical technique is attached to the original method parsing method. 2) The message digest and the final message digest formula was enhanced by giving additional shifting, xoring and improved calculations. The enhanced SHA-1 maintains its original rounds which consist of 80 rounds and message digest output of 160 bit. To test the enhanced SHA-1, the author entered the same value as input message to the original and enhanced algorithm and based from the figures on table 5 both algorithm have 160 bit message digest output but the enhanced SHA-1 produces a different value. The table 7 shows a minimal delay on the processing speed, this is expected due to some processes incorporated on the enhanced algorithm. On table 6, the enhanced algorithm produces a very evident avalanche effect which is considered a good cryptographic design. The enhanced algorithm is better and more secured version of SHA-1 which is expected to resist possible future collision attacks.

-Digital-Signature-Scheme-for-EmailMessages.html [5]

Schneier B., Schneier Security. Article: Cryptanalysis of SHA-1, https://www.schneier.com/blog/archives/2 005/02/cryptanalysis_o.html

[6]

Alam M., and Ray S., Design of an Intelligent SHA-1 Based Cryptographic System: A CPSO Based Approach. International Journal of Network Security, Vol. 15, No.6,PP.465-470, Nov. 2013

[7]

Wang X., and Yu H., “How to break MD5 and other hash functions”, Advances in Cryptology – EUROCRYPT, LNCS 3494, Springer-Verlag , pp.19-35, 2005.

[8]

Goldwasser S., Micali S., and Rivest R., “ A Digital Signature Scheme Secure Against Adaptive Chosen-Message Attacks”, Journal on Computing vol. 17, no.2. pp281-308, 1988.

[9]

“New Europian Schemes for signatures, Integrity and Encryption Project”, available at: http://www.cryptonessie.org.

[10]

Rjasko M., Properties of Cryptographic Hash Function, https://www.fmph.uniba.sk/fileadmin/user _upload/editors/studium/svk/2008/INF/rja sko.pdf

[11]

The SHA-1 attacks, http://en.wikipedia.org/wiki/SHA-1

REFERENCES [1]

[2]

Alfred M., Oorschot P., and Vanstone S., Handbook of Applied Cryptography, CRC Press, 1997 Lakshmanan T., and Muthusamy M., A Novel Secure Hash Algorithm for Public Key Digital Signature Schemes. The International Arab Journal of Information Technology Vol.9, No.3 pp. 262 – 267, May 2012

[3]

Rhee M., Internet Security, Cryptographic principles, algorithms and principles, pp.149, 2003 John Wiley & Sons, Ltd ISBN 0-470-85285-2

[4]

Yang H., Digital Signature Scheme for Email Messages. PKI Tutorials. http://www.herongyang.com/PKI/SMIME

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Christine Charmaine G. San Jose is currently taking up her Doctor in Information Technology and now on dissertation writing at Technological Institute of the Philippines, Quezon City, Philippines and finished her MSIT degree at

8


University of La Salette, Santiago City, Philippines in the year 2005. She obtained her Bachelor of Science in Computer Science degree at Emilio Aguinaldo College, Manila, Philippines in the year 1998. She started her teaching profession in the year 2001 at Institute of Information and Communication Technology at Isabela State University, Echague, Isabela, Philippines. She has been designated as an Institute Secretary for four years and has been extensively involved in Research and Extension of the Institute. She has been tapped by various Agencies to perform Job relevant to her field of specialization. She has been a Board of Election Inspector (BEI) of Department of Science and Technology (DOST) and was designated by the Commission on Election (COMELEC) to be a Member of the Board of Canvasser as Consolidated Canvassing System Operator. She is a member and an officer of JCI – Junior Chamber International Philippines. Her field of interest includes information system and data security.

Dr. Bobby D. Gerardo is currently the Vice President for Administration and Finance of West Visayas State University, Iloilo City, Philippines. His dissertation: Discovering Driving Patterns using Rule-based intelligent Data Mining Agent (RiDAMA) in Distributed Insurance Telematic Systems. He has published more than 60 research papers in national and international journals and conferences. He is a referee to international conferences and journal publications such as in IEEE Transactions on Pattern Analysis and Machine Intelligence and IEEE Transactions on Knowledge and Data Engineering. He is interested in the following research fields: distributed systems, telematics systems, CORBA, data mining, web services, ubiquitos computing and mobile communications.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Dr. Bartolome T. Tanguilig III took his Bachelor of Science in Computer Engineering in Pamantasan ng Lungsod ng Maynila, Philippines in 1991. He finished his Master’s Degree in Computer Science from De La Salle University, Manila, Philippines in 1999, and his Doctor of Philosophy in Technology Management from Technological University of the Philippines, Manila in 2003. He is currently the Assistant Vice President for Academic Affairs and concurrent Dean of the College of Information Technology Education and Graduate Programs of the Technological Institute of the Philippines, Quezon City. Dr. Tanguilig is a member of the Commission on Higher Education (CHED) Technical Panel for IT Education, the chair of the CHED Technical Committee for IT, the founder of Junior Philippine ITE Researchers (JUPITER), board member of the Philippine Society of IT Educators (PSITE), member of the PCS Information and Computing Accreditation Board (PICAB), and a member of the Computing Society of the Philippines (CSP).

9


Grading of Carabao Mango Using Image Processing Lorena C. Ilagan1, Jerry V.Turingan1, Alexiel K. Aranas2, Emmanuelle D. Ignacio2, Elbert C.Rasay2 Faculty of Engineering, University of Perpetual Help System DALTA1 and Department of Electronics Engineering, University of Perpetual Help System DALTA2 ABSTRACT Grading the Quality of the mango is usuallyhandled physically which might not beaccurate due to the judgement of a person. Inthis study, image processing using MATLABis proposed using MandamiFuzzyInterference System. The Mango is placed onchamber to help the camera in collecting theRed-Green-Blue (RGB) data and the spotdata of the program. This woodenrectangular chamber includes three LEDflashlights and the camera. The resultsshowed that Red mean value less than orequal to 155 are acceptable. The blemishesand maturity of the mango is then analysedbased in the white spots from the image andred means of the image. The paper alsoincludes size detection using boundarytracing and spot detection using the regionprops function. The results show that theprogram is accurate enough to grade themango. The system is strict enough to rejectthe sample mango if it is too large, is toosmall, is yellow or is blemished.

KEYWORDS Image processing, Edge Detection, RGB values, Carabao Mango, Mamdami Fuzzy Inference System

mango is the grading of the quality of the mango which is normally handled physically as most of the criteria in judging the quality of the mango is usually based on the colour which is usually easier to detect by the use of human senses. It is possible to grade each individual mango through image processing. Image processing is important in detecting the colour and the height of theCarabao mango as well as detecting the most common deficiencies mangoes may acquire while maturing. The image processing will be monitored by the use of a MATLAB program. This study can be beneficial to agricultural labour as the program may independently and accurately grade the quality of each individual mango. It may also help lessen the amount of time in the process of mango grading. As it grades Carabao mangoes, it will produce packages of more graded and quality Carabao mangoes to deliver. It is also beneficial to different companies that endorse the transportation and export quality Carabao mango products and productions as it will standardize the criteria for grading Carabao mango.

1 INTRODUCTION Mango is one of the most common tropical fruit produced worldwide. It is even regarded in the Philippines as its national fruit. As such, it is very important that the mango must be properly treated in its production, its harvest and its delivery. One of the most important processes in handling

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

2 REVIEW OF RELATED STUDIES Many studies have shown that it is already possible to grade mango with the use of different sensors. However, these studies focus more on the maturity of the mango rather than the quality of the mango itself. These studies rely on having an external

10


sensor to gather the data instead of having a camera to directly feed the data to the computer program. The studies are also done with a different breed of mango, which means, their data may be different from the data we have gathered. A research headed by Ms.Dadwal can “Estimate Ripeness Level of fruits Using RGB Color Space and Fuzzy Logic Technique” This research checks for the ripeness of any fruit by collecting the means of the RGB values of the given fruit through colour segmentation then grade the ripeness of the fruit by the use of the Fuzzy Logic.[1] A research headed by Mr.Mansor proposes a “Fuzzy Ripening Mango Index Using RGB Sensor Model”which uses a RGB colour sensor to collect the data which will be graded by the fuzzy system into three classes, which are based on maturity mangoes by detecting the distinct colour of the sample mangoes. This research concludes with an accuracy of 85% in its segregation.[2] Mr.Nandi and his group of researchers has made a “Machine Vision Based Techniques for Automatic Mango Fruit Sorting and Grading Based on Maturity Level and Size” which relies on image processing and edge detection to be able to segregate different samples of Indian mango by the use of Fuzzy logic algorithm. The data is collected with a CCD camera sensor which makes it easier to gather the RGB values of the mango.[3] A researchheaded by Mr.Razak, which is “Mango grading by using Fuzzy Image Analysis”, uses image processing to collect the RGB values of the image and its height and utilize the Fuzzy image analysis to produce the output to different grades. This research yielded an 80% in accuracy. [4]

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

3 METHODOLOGY A chamber is produced to help aid the camera in collecting the RGB data and the spot data of the program. This wooden rectangular chamber includes three LED flashlights and the camera. By providing enough lights, the camera will capture the minimum amount of shadows that can affect the data collected. A funnel is provided in case the mango to be sampled needs to be shot upright without interfering in the view of the camera. From the collected image, the computer program will collect the height and diameter of the sample mango by the use of boundary tracing. From the traced boundary, the program will measure the distance between one boundary to the other and gather the largest X and Y distance for the height and diameter respectively.

Figure 1.Edge Detection for Height Detection

The program will produce the amount of black spots by subtracting the collected amount of white spots from the maximum number of white spot,

Where:

B = Wmax – W picture

(1)

Wmax =

number maximum pixels

of white

Wpicture =

number of white pixels in the picture

11


and this information will subtract the product between the height and diameter of the sample object. Area = X*Y Where:

(2)

X=

Height of mango

Y=

Diameter of mango

Spot = Area - B

After collecting the values of the segmented picture, the red values will be averaged.

(4)

The difference will be divided to the product of the dimension of the sample object; therefore, a sample that produces a 100% output means the system does not detect any spot from the sample object. P = Spot/Area

value of the mango to ensure that the colour of the background does not interfere in calculating for the average of the collected value of the mango.

(3)

Three data, the height of the mango, the percentage of spots and the red mean value, will be inputted in the Fuzzy Inference system to be able to grade whether the sample mango is acceptable for export or will be rejected instead.

Percentage is used to see how much of the mango is already blemished as compared to counting the amount of blemish in the mango.

Figure 3.Fuzzy Inference System and its members Figure 2. Spot Detection using Regionprops function

The maturity will be based on the red means of the sample object. As the target colours are Green and Yellow, the only value that varies is the Red pixel values as a high red pixel value, a high green pixel value and a low pixel value produces a yellow colour and a low red pixel value, a high green pixel value and a low pixel value produces a green colour. The program will start by segmenting the colour between the mango and the background before collecting the

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 4.Membership function of Height variable

12


4 DATA, RESULTS AND DISCUSSION

Figure 7. Sample Object Figure 5. Membership function of Spot variable

A sample mango (Fig 8) which was 12 cm. tall was used to scale the distance between pixels as well as to check the limit of the height of the mango in pixels. From this mango sample, a 533 px tall mango is produced. Calculating for the limits of the acceptable height of an exportable mango, the for the lowest acceptable height of the mango, which is 10.5 cm, is scaled to 466 px; the highest acceptable height of the mango, which is 14 cm, is scaled to 622 px.

Figure 6.Membership function of Maturity variable

There are five rules in the Fuzzy Inference system -

If height(X) is Large, then the sample is Rejected If height(X) is Small, then the sample is Rejected If percentage (P) is visible, then the sample is Rejected. If color is yellow, the sample is Rejected If height(X) is Enough and the percentage (P) is invisible and colour is Green, then the sample is Accepted.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 8. An unripe mango and a ripe mango.

Two sample mangoes(Fig 9), one of which is unripe for a few days, and another which shows yellow skins on both the apex area and the stalk area, is used to calculate for the maturity of the mango. The former produces a mean red value of 155 and the latter produces a mean red value of 173. After gathering the digital data to set the limits for the Fuzzy Inference System, thirty mango samples were subjected for testing.

13


Spot (percent) 0.9747 0.9929 0.9932 0.9980 0.9673 0.9715 0.9893 0.9912 0.9703 0.9961 0.9886 1 0.9656 0.9557 0.9954 0.9953 0.9879 0.9951 0.9940 0.9889 0.9918 0.9887 0.9954 0.9934 0.9917 0.9764 0.9903 0.9997 0.9819 0.9972

Output (Accept/Reje ct)

450 473 565 631 492 502 635 622 611 592 635 638 465 479 593 639 503 501 490 613 506 519 627 630 572 482 395 637 437 489

Maturity (Red means in pixel)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Length (pixel)

Sample

Table 1 Input of the samples and the output from the Fuzzy Inference System

157 166 0 166 161 158 160 0 154 0 0 0 152 0 174 153 0 166 169 148 0 164 159 145 0 157 147 161 151 170

Reject Reject Accept Reject Reject Accept Reject Accept Accept Accept Reject Reject Reject Reject Reject Reject Accept Reject Reject Accept Accept Reject Reject Reject Accept Accept Reject Reject Reject Reject

The results show that the program is accurate enough to grade the mango. The system is strict enough to reject the sample mango if it is too large, is too small, is yellow or is blemished. While the system is more lenient on the grading between the spot detection considering our reference shows that there must be 0% tolerance in the spot detection of the system, it still show that the system may not be perfect as the

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

camera can still detect shadows inside the object.

5 CONCLUSION In this paper, the maturity of the Carabao Mango was collected by gathering the data of the RGB value through color image segmentation algorithm. The results showed that Red mean value less than or equal to 155 are acceptable. The paper also includes size detection using boundary tracing and spot detection using the regionprops function. Mamdani-type Fuzzy Inference System was used to determine the quality of the Carabao Mango from the data gathered by the aforementioned image processing techniques. Although the parameters can be easily adjusted by manipulating the values in the Fuzzy Inference System, with the given references and data, the program is successful in grading the mango whether it is acceptable enough for export through three different parameters. ACKNOWLEDGEMENT UNIVERSITY OF PERPETUAL HELP SYSTEM DATA

REFERENCES Tajul Rosli B. Razak, Mahmod B. Othman, Mohd Nazari bin Abu Bakar, Khairul Adilah bt Ahmad, Ab Razak Mansor. “Mango Grading By Using Fuzzy Image Analysis.” International Conference on Agricultural, Environment and Biological Sciences 2012.. pp. 18-22. [1]

14


[2] Dadwal, Meenu and Banga, Vijay Kumar, “Estimate Ripeness Level of fruits Using RGB Color Space and Fuzzy Logic Technique.” International Journal of Engineering and Advanced Technology 2012, pp. 225-229. [3] Mansor, Ab Razak, Mahmod Othman, Khairul Adilah Ahmad and Tajul Rosli Razak. “Fuzzy Ripening Mango Index Using RGB Sensor Model.” ResearchersWorld-Journal of Arts, Science & Commerce. Vol.–V, Issue–2, April2014

[4] Nandi, C. S., Tudu, B. and Koley, C. “Machine Vision Based Techniques for Automatic Mango Fruit Sorting and Grading Based on Maturity Level and Size.” Switzerland : Springer International Publishing, 2014.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

15


Hardware Architecture of FAST Algorithm for Feature Point Detection Chang-Sue Seo1, Hoon-Ju Chung1, Sung-Young Kim1, Sangook Moon2 and Yong-Hwan Lee1 Kumoh National Institute of Technology, Korea1 and Mokwon University, Korea2 [email protected], [email protected], [email protected], [email protected] and [email protected]

ABSTRACT In this paper, we present method that detects useful feature points based on hardware architecture. We propose hardware architecture that uses the algorithm of FAST-n[1]. Feature point detection process needs extensive computing power and processing time. Therefore, we build hardware structure for real-time processing. The structure of the hardware is as follows. After loading the images in parallel, finding feature point candidates and selecting valid feature point modules operate simultaneously and independently using pipeline structure to reduce processing time. Proposed hardware architecture will operate in about 20,000 cycles in case of 320 x 240 resolution image. If our hardware structure is used for 1080p, the performance of processing will be about 70fps.

Therefore, we propose hardware structure for real-time processing application. The hardware implementation is faster and requires fewer resources than software structure. The structure of the hardware is as follows. After loading the images in parallel, three modules which search the feature point candidates module, compute score of each feature point and select effective feature point in the feature point candidates module operate simultaneously and independently using pipeline structure. As a result, faster operation than software can be achieved.

2 FAST Algorithm

KEYWORDS FAST, Corner detection, Edge detection, Feature point, Hardware architecture

1 Introduction

Figure 1. Segment test corner detection in an image patch

Task of extracting feature points is the first step of many vision tasks such as object tracking, SLAM (simultaneous localization and mapping), localization, image matching and recognition. To extract feature points, a number of algorithms have been studied[1-7]. Most of the algorithms require extensive computational cost and time. These methods are not suitable for real-time processing of images.

FAST algorithm is abbreviation of features from accelerated segment test. It loads 16 pixels around the circular for a single reference pixel P as shown Figure 1. At this time, as formula 1, each pixel is compared if it is greater than plus the threshold value to the reference pixel, or if it is smaller than the reference pixel minus the threshold value. The result of comparison will be divided into three states, point darker than

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

16


the reference pixel, similar point and point brighter than the reference pixel.

Decision tree scheme is used for detecting continuous pixel, as shown in the Figure 3 [8].

(1) The number of continuous dark or bright pixels determines the constant n of FAST-n algorithm. The FAST-n algorithm usually detects 7 to 12 consecutive pixels.

Figure 3. Decision tree of dark point, similar point, bright point

When this condition is satisfied, the reference pixel is selected as the feature point candidate. In this case, detected feature point candidates are often located around detected feature points because of characteristics of FAST algorithm. However FAST algorithm requires postprocessing called NMS (non-maximal suppression). NMS selects valid feature points among the detected feature point candidates. Through formula 2, the maximum threshold that meets the conditions of the feature point candidate is calculated and stored. There are several intuitive definitions for V: 1. The maximum value of n for which p is still a corner. 2. The maximum value of t for which p is still a corner. 3. The sum of the absolute difference between the pixels in the contiguous arc and the center pixel. Definitions 1 and 2 are very highly quantized measures, and many pixels share the same value of these. For speed of computation, a slightly modified version of 3 is used. V is given by : [1]

Figure 2. Center pixel and neighbor pixels (top: pixel value, middle: bright, bottom: dark)

Figure 2 illustrates how to classify pixel. The around pixels are compared to the value of center pixel plus threshold value 64. Then the around pixels have the value of 0 or 1 according to the differences.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

(2)

To detect a valid feature points, only the feature point candidate which has the maximum value compared with others are left and the rests are removed.

17


Figure 4. Result Einstein image of FAST-9 algorithm in case of different threshold (top 16, middle 32, bottom 64)

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 5. Result Lena image of FAST-9 algorithm in case of different threshold (top 16, middle 32, bottom 64)

18


Figure 4 and 5 are the result of FAST-9 algorithm simulation in Matlab. Size of the image is 320 x 240. Green circles are feature points candidate and red dots are NMS processed feature points.

Figure 6 illustrates the count of feature point candidate and NMS processed feature point.

Table 1. Result from Lena image of FAST-9 algorithm Threshold FD NMS per Threshold FD NMS per Threshold FD NMS per Threshold FD NMS per

8 6,456 2,654 41.11 40 431 249 57.77 72 70 45 64.29 104 12 11 91.67

16 2,732 1,239 45.35 48 267 160 59.93 80 49 33 67.35 112 8 7 87.50

24 1,418 704 49.65 56 171 105 61.40 88 27 22 81.48 120 6 5 83.33

32 777 419 53.93 64 120 73 60.83 96 19 17 89.47 128 3 3 100.00

Table 1 illustrates the result of FAST-9 algorithm applied to Lena image with varying threshold from 8 to 128. FD is number of feature point before NMS processing. NMS is number of feature point after NMS processing. Per is the percentage of NMS processed feature point number of feature point candidate number. It is shown that as number of feature point decrease as threshold increase.

Figure 7. Result Lena image of FAST-9 algorithm in case of different threshold (top 16, middle 32, bottom 64)

Figure 7 illustrates the percentage of NMS processed feature point number of feature point candidate number. It is shown that as threshold increase, so do percentage.

3 Hardware Structure of FAST Algorithm The Figure 8 illustrates the structure of software briefly. First, the image to gray scale is performed. Then FD function, FS function, NMS function are called in sequence.

Figure 8. Software structure of FAST-n

Figure 9 illustrates overall hardware structure. Divided gray scale image is stored in block RAM. Then pipeline structure operates with the stages of FD, FS and NMS modules simultaneously and independently. Figure 6. Result Lena image of FAST-9 algorithm in case of different threshold from 8 to 128

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

19


P_ref

+

cmp

C[0] th

S_

cmp

C[1]

bright

... cmp

C[15]

is_FD

P_ref

-

cmp

C[0] th

S_

cmp

C[1]

dark

... cmp

C[15]

Figure 11. Block diagram of FD (feature detection) module Figure 9. Hardware structure of FAST-n

When the center pixel and the neighbor pixels in the block RAM are loaded, neighbor pixels are made to vector as shown in Figure 10 so that vector is processed easily in hardware.

Figure 12 illustrates block diagram of FS (feature score) module. Score is calculated using bmin and bmax. That module requires from minimum 0 to maximum 8 cycles to find accurate score. However, because overall system takes much time due to this operation, we changed the architecture from repetitive FS to a series of FS through pipeline. Flip flop is added to series of FS in each FS module as shown Figure 13. P_ref

+ C[0]

cmp

th C[1]

cmp

S_ bright

... C[15]

cmp

P_ref

C[0]

cmp

th C[1]

cmp

S_ dark

...

Figure 10. Aligned arc pixel

Figure 11 illustrates block diagram of FD (feature detection) module. Adder and subtractor are used to compute values from pixel and threshold. Values are compared with values in vector. If those results consist of continuous 1s of which the number is more 9, reference pixel would be a feature point candidate.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

C[15]

0 1

bmin

bmax

0 1

cmp

+

>>1

0

0 1

==?

Figure 12. Block diagram of repetitive FS (feature score) module

20


P_ref

+ C[0]

cmp

th C[1]

cmp

S_ bright

... C[15]

cmp FF

P_ref

C[0]

cmp

th C[1]

cmp

S_ dark

... C[15]

0 1

bmin

bmax

cmp

+

0 1

0 1

>>1

0

FF

ACKNOWLEDGEMENT

==?

Figure 13. Block diagram of revised FS (feature score) module

Figure 14 illustrates the block diagram of NMS (non-maximal suppression) module. Calculated reference pixel score is stored to memory and compared with adjacent pixel score. If reference pixel score is the highest, reference pixel would be a valid feature point. P_ref_score C_score [x-1,y-1] C_score [x-1,y] C_score [x-1,y+1] C_score [x,y-1]

cmp cmp cmp

S_ bright

==?

resolution image. If our hardware structure is used for 1080p, the performance of processing will be about 70fps. Object tracking is a key component in the system of caring companion animals. Proposed method would play a key role for efficient and fast tracking in the video module of the caring system.

This work (Grants No. C0217661) was supported by Business for Cooperative R&D between Industry, Academy, and Research Institute funded Korea Small and Medium Business Administration in 2014.

REFERENCES [1]

E. Rosten and T. Drummond, “Machine learning for high-speed corner detection,” European Conference on Computer Vision, vol. 3951, pp. 430-443, May 2006.

[2]

D. Lowe, “Distinctive Image Features from ScaleInvariant Keypoints,” International Journal of Computer Vision, vol. 60, pp. 91-110, November 2004.

[3]

C. Harris and M. Stephens, “A Combined Corner and Edge Detector,” Alvey Vision Conference, vol. 15, pp. 147-151, 1988.

[4]

H. Bay, T. Tuytelaars, and L. Gool, “SURF: Speeded-Up Robust Features,” European Conference on Computer Vision, vol. 3951, pp. 404-417, May 2008.

[5]

T. K. Kim, “An Embedded FAST Hardware Accelerator for Image Feature Detection,” Journal of The Institute of Electronics Engineers, vol. 49, pp. 28-34, March 2012.

[6]

S. R. Kim, H. J. Yoo, and K. H. Sohn, “FAST and BRIEF based Real-Time Feature Matching Algorithms,” The Korean Society of Broadcast Engineers, vol. 2012, pp. 1-4, November 2012.

[7]

H. Heo and K. Y. Lee, “FPGA based Implementation of FAST and BRIEF algorithm for Object Recognition,” Journal of IKEEE, vol. 17, pp. 202-207, June 2013.

[8]

J.R. Quinlan, “Induction of decision trees,” Machine Learning 1, vol. 1, pp. 81-106, 1986.

is_corner

0

cmp

Figure 14. Block diagram of NMS (non-maximal suppression) module

4 Conclusions We proposed hardware architecture of FAST algorithm for real-time processing. The pixel of gray scale image are divided and stored in block RAM. Pipeline structure is applied to FD (Feature Detection) module, FS (Feature Score) module and NMS (Non-Maximal Suppression) module in order to operate simultaneously and separately. Proposed hardware architecture will operate in about 20,000 cycles in case of 320 x 240

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

21


A Small Area 5GHz LC VCO with an On-Chip Solenoid Inductor using a 0.13μm Digital CMOS Technology Chul Nam1, Byeungleul Lee2, Tae-Young Byun3, Yongjun Jon4, and Bonghwan Kim5,* 1 R&D Center/Siliconharmony, Seong Nam-Si, Korea 2 Mechatronics Engineering, Korea University of Technology and Education, Chungnam, Korea 3 School of Information Technology Engineering, Catholic University of Daegu, Gyeongbuk, Korea 4 TIOS Inc., Gyeongbuk, Korea 5 Electronics Engineering, Catholic University of Daegu, Gyeongbuk, Korea *Email:[email protected]

ABSTRACT This paper presents a small area LC VCO with an on-chip solenoid inductor using 0.13μm digital CMOS technology. This on-chip solenoid inductor is vertically built by metal and via layers with a horizontal occupation which gives the advantages of small area due to 3-D nature compared to a spiral inductor. Based on 3-D EM simulation and analysis, the unified equivalent model for a solenoid inductor is proposed so that it could be adapted for LC Voltage controlled oscillator in the digital standard CMOS process. The designed LC VCO has the occupation area as 0.013 mm2 and the frequency tuning range as 1GHz. The measured phase noise of VCO was measured as -105.25dBc/Hz at a 1MHz offset with the power consumption of 13.2mW with total area of 0.028mm2.

KEYWORDS On-chip solenoid simulation, CMOS

inductor,

LC

VCO,

EM

1. INTRODUCTION Even though the high speed digital complementary metal-oxide semiconductor (CMOS) technology has been developed up to several GHz in the frequency, the inductorcapacitor (LC) voltage controlled oscillator (VCO) are still requiring the RF process with the ultra-thick metal spiral inductor. In other hands, for the sake of low cost process, the

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

digitally controlled oscillator (DCO)s have been studied operating at the several gigahertz range in the digital CMOS technology [1-2]. The spiral inductor used in DCOs, however, is still a copy of one used in RF process sacrificing quality factor and occupying large area, what is worse, causing the high processing cost. In addition, circuit designers have to face the challenges in designing one’s own spiral inductor and take responsibility for its performance even after electromagnetic (EM) simulation was done through EDA tools such as Agilent Momentum, FEMLAB, Sonnet, ASTIC [3] and HFSS [4] etc. A study of solenoid inductor has been done mainly in microelectro-mechanical-systems (MEMS) technology [5]. An electroplated MEMS inductor had been reported with a high quality factor and an inductance of a tenth of nH owing to the good conductivity of copper [6]. However, its application into CMOS process results in the additional mask steps and increases the processing cost. Before 0.13μm CMOS process, the poor conductivity of Al in CMOS process had to use a thick metallization method interconnecting more than two metals to reduce the series resistance of inductor. Therefore, the copper metallization below 0.13μm CMOS standard process has introduced the possibility of implementing a low series resistance inductor in LC type VCO while the large area spiral inductor is still obstacle for the low cost standard CMOS process.

22


In this paper, on-chip solenoid inductors using metal and via stacking were designed and their electrical characteristics in terms of inductance and quality factor were simulated using 3D-EM simulator [7]. Designed inductors according to the different dimension had been implemented and measured its characteristics by two port Sparameters using a HP E4440A. The measured S-parameters were used in analyzing the proposed equivalent lumped model and further the linearized equation through the parameter optimization. Based on this model, a LC type VCO with on-chip solenoid inductor had been designed and measured in regards to its tuning range and phase noise.

2. ON-CHIP SOLENOID INDUCTOR

number of turns (N) and the pitch (p) between each turn. As the cross area is determined by the process technology, the inductance is then proportional to the width (W) and the number of turns (LS ∝ N∙W). Normally, the series resistance of the inductor (RS), which mainly affects the quality factor, is increased by the skin effect [8] and the proximity effect according to operating frequency. In solenoid inductor, the proximity effect can be ignored as the conductor moves farther from the adjacent turns. RS is rewritten approximately as;

𝑅𝑠 = 𝑅𝑑𝑐

𝑡 𝛿 ∙ (1 −

−𝑡 𝑒𝛿)

= 𝜌□ ∙

δ=

Usually, the magnetic flux of the spiral inductor using the planar CMOS process penetrates the substrate with its axis perpendicular to the surface. Naturally, the spiral inductor has impairment about a low quality factor due to the substrate loss and a reason to put the pattern ground shield under the inductor. On the other hand, the building of solenoid inductor is completed to connect the top plates (M6) and the bottom plates (M1) through vias (V1-V5) as shown in Fig. 1 [7]. Since its axis is parallel to the substrate, the solenoid inductor is less susceptible to the substrate losses due to the eddy current.

𝜌 𝜋𝜇𝑓

𝑡 −𝑡

𝛿 ∙ (1 − 𝑒 𝛿 )

(1)

(2)

where t, Ww and Wl represent the metal thickness, the post width and the post length, and ρ, μ and ƒ are the resistivity of the copper, the permeability of the air and the operating frequency, respectively [7]. For 5GHz frequency applications, the skin depth, δ, is about 0.94μm and can be ignored as it is larger than the top metal thickness 0.9μm, for example, in 0.13μm digital CMOS process[7]. The series resistance, RS becomes 1.2∙Rdc and in the same way, the quality factor can be expressed as Q = 0.833

Top Plate

2𝑁𝑊 + 2𝑊𝑤 𝑊𝑙

𝜔 ∙ 𝐿𝑆 1 ≈ 𝑅𝑑𝑐 𝑊∙𝑁

(3)

PORT2 ℓ

p

M6 M5 M4 M3 M2 M1

Wl 0.37µm

H

W Ww W

Bottom Plate PORT1

V5 V4 V3 V2 V1

0.28µm

Figure 1. A structure of the solenoid inductor [7].

In general analysis, the macro-scale inductance of solenoid is expressed in terms of its dimension such as cross-section area, the

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

In equation 3, the width (W) is quite larger than the post width (Ww) and length (Wl) and Rdc is also proportional to N∙W, and then the quality factor becomes inversely proportional to N∙W. From equation (1), (2) and (3), we can suggest the design parameters of the solenoid inductor as W and N. These parameters give the good golden rule designing the solenoid inductor.

23


the varactor capacitance at nth control signal can be expressed in terms of the effective switchable capacitance and control signals as;

3. LC VCO DESIGN & CHARACTERIZATION The proposed LC VCO with on-chip solenoid inductor consists of a negative-gm circuit, a pair of varactor for fine tuning, and a varactor tuning bank for coarse tuning as show in Fig. 2 and Fig. 3. Once the solenoid inductance (LS) is determined according to the number of turn (N) from the linearized equation [7], the oscillation frequency of VCO can be calculated with the fine and coarse tuning capacitance by; Equation 4. 𝑓𝑜𝑠𝑐 =

2𝜋 ∙ 𝐿𝑠 ∙ (𝐶𝑑𝑠𝑣 + 𝐶𝑓𝑣𝑎𝑟 ) N*RS

(4)

N*LS P2

N*CP CSI

N*CP RSI

(5)

Furthermore, if varactor capacitor array is binary-weighted, the total capacitance of the varactor bank becomes; 𝐶𝑑𝑠𝑣 = 23 − 1 ∙ 𝐶𝑙𝑜𝑤 ,0 + ∆𝐶𝑑𝑠𝑣 𝑛=2

∆𝐶𝑑𝑠𝑣 = ∆𝐶𝑑𝑠𝑣 ,0

𝑛=0

1

P1

RSI

𝐶𝑑𝑠𝑣 ,𝑛 = 𝐶𝑙𝑜𝑤 ,𝑛 + 𝐷[𝑛] ∙ ∆𝐶𝑑𝑠𝑣 ,𝑛

2𝑛 ∙ 𝐷[𝑛]

(6) (7)

Then, the maximum variable capacitance which contributes the total frequency range is 7 times the effective switchable capacitance, that is ΔCdcv_max= 7·ΔCdsv,0=7·(Chigh,0-Clow,0), in which the subscript 0 denotes LSB. This makes the tunable frequency range as;

CSI

Figure 2. The proposed unified equivalent model of solenoid inductor.

𝐶𝑙𝑜𝑤 ,0 ∆f = 𝑓𝑚𝑎𝑥 − 𝑓𝑚𝑖𝑛 = 𝑓𝑚𝑎𝑥 (1 − 𝐶ℎ𝑖𝑔ℎ,0

𝑓𝑚𝑎𝑥 ≅

1 2𝜋𝐿𝑠 ∙ 7 ∙ 𝐶𝑙𝑜𝑤 ,0

, 𝑓𝑚𝑖𝑛 ≅

1 2𝜋𝐿𝑠 ∙ 7 ∙ 𝐶ℎ𝑖𝑔ℎ,0

(8) (9)

where, Cfvar < 7·Clow,0 D (1)

: otherwise where  is a adjusting factor.

Figure 2. Round organization in D-LEACH

Z. U. A. Jaffri, et al, proposed a new clustering routing protocol, zone and energy threshold based clustering routing protocol that is application-aware and heterogeneity-aware. In the execution environment, this distributed algorithm divides the whole network area into multiple zones. After dividing the nodes into multiple zones, it selects a cluster head node with the highest energy-efficiency in each zone [13]. 3 PROBLEM DEFINITION As mentioned earlier, ideal percentage of cluster heads is 5% of all nodes for effective energy-efficient operation in LEACH [1]. However, it is not tend to maintain 5 % of cluster heads because cluster heads are determined with probability function in distributed manner at each sensor node. By computer simulation, we identified that percentage of cluster heads vary from 4% to 6% is about 40% in all rounds or whole network lifetime in LEACH. As cluster size increases, that is, the number of clusters decrease, energy consumption for communications between noncluster heads and a cluster head also increases. Consequently, there may be great difference in residual energy among cluster head and noncluster heads. On the contrary to this, as the number of clusters increase, energy consumption between cluster heads and sink

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

node increase. Especially, communication cost between cluster head and sink node is very high compared to the communication cost between cluster head and non-cluster head in cluster. So, it is needed to balance residual energy of all sensor nodes across whole network lifetime. LEACH and D-LEACH method of determining cluster head using probability function poses some limitations in balancing residual energy. 3.1 Imbalance of the Amount of Energy Consumption with the Change of Disparity of Cluster Size In Figure 3(a), there is only one cluster by LEACH scheme, and cluster size with cluster head CH42 is very high. In this case, energy consumption for interior communication between non-cluster heads and cluster head CH42 can comparatively be greater than that of another cluster. In Figure 3(b), there are eight clusters, and we identify that there are the wide disparity in cluster sizes. For example, cluster size with cluster heads CH34 is 19. On the contrary, cluster size with cluster head CH88 is only 3. This case can lead to the imbalance of energy consumption among nodes in the field. Examples of cluster shapes formed by DLEACH are shown in 3(c) and 3(d). In DLEACH, clustering is also similar to that in LEACH, so there can be the imbalance of energy consumption among nodes in the field.

(a) Example of cluster shape in case the number of

clusters is only one in LEACH (N=100, P=0.05, K=5, D=20) Figure 3. Examples of cluster shape in LEACH and DLEACH(continued)

111


(b) Example of cluster shape in case that the number of

clusters is eight in LEACH (N=100, P=0.05, K=5, D=20)

To show disparity of the number of clusters and cluster size per round, we analysed average values and standard deviation of both the number of clusters and cluster size in LEACH and D-LEACH. We assumed that total number of sensor nodes across WSN is 100. Both the average number of clusters and cluster size gradually decrease when some nodes disappear due to exhaustion of energy. Especially, when the ratio of live nodes is below 50 % such as (C), average number of clusters decreases sharply. Average cluster size in D-LEACH is more balanced than that of LEACH. These imply that the number of live nodes per round affects cluster formation in next round. Table 1. Comparison of cluster properties between two schemes

(c) Example of cluster shape in case that the number of

clusters is three in D-LEACH (N=100, P=0.05, K=5, D=20)

Number of live nodes Metrics

100 (A)

65~99 (B)

30~65 (C)

0~29 (D)

Avg. number of clusters

5

4.2

2.6

1.1

Std. dev. of clusters

1.8

1.8

1.1

1

Avg. number of clusters

5

4.2

2.8

1.1

Std. dev. of clusters

1.7

1.7

1.3

0.7

19

17.9

16.8

8.3

Std. Dev. of Cluster Size

11.1

9.7

9.3

5.6

Avg. Cluster Size

18.7

18.7

17.1

11.4

Std. Dev. of Cluster Size

10.7

10.8

9.6

5.4

Clustering scheme

LEACH

D-LEACH

Avg. Cluster Size

LEACH

D-LEACH

4 CENTRALIZED CLUSTERING SCHEME 4.1 Key Ideas (d) Example of cluster shape in case that the number of clusters is seven in D-LEACH (N=100, P=0.05, K=5, D=20) Figure 3. Examples of cluster shape in LEACH and DLEACH

3.2 The Number of Clusters per Round

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

In this paper, we extend previous study of centralized cluster head selection scheme [8] which introduced a novel clustering scheme, called HD-LEACH (hybrid density-based LEACH). We describe details of HD-LEACH, analyze the amount of energy consumption

112


mathematically, and compare important performance results with existing schemes considering cluster formation and network lifetime. We assume that there is a special node, base station (BS) to select cluster heads among all sensor nodes in centralized manner. We assume that BS can be a network device installed in the center of field. Also, a proper BS can be elected by calculation of relative location between a candidate BS and other nodes automatically. BS gathers information about amount of residual energy of each node from all nodes before entering next round. Then, BS stores the information of residual energy in a table, and sorts amount of residual energy of each node in descending order. Finally, BS selects 5% nodes with higher amount of residual energy as cluster heads for next round. The number of selected cluster heads can be calculated as follows

K   N  P

(2)

where K is the number of cluster heads, N is the number of live nodes in current round, and P is percentage of cluster heads to all nodes. Here, we set P to 0.05(5%). If N and P are 100 and 0.05 respectively, K can be 5. As round progresses in network lifetime, N can be changed because some nodes die due to energy exhaustion. Then, K also can be changed to reflect N, the number of live nodes at current round. HD-LEACH is based on existing DLEACH, but cluster heads are selected by a special base station that is responsible to determine proper cluster heads prior to current round in centralized manner. Each round in HD-LEACH is similarly organized as that of DLEACH (Figure 4).

Figure 4. Round organization of HD-LEACH

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

We illustrate an operational procedure of HDLEACH in Figure 5. There are a CH and non-cluster head nodes in a cluster, and cluster membership is decided by base station in centralized method that differs from LEACH and D-LEACH. In pre-clustering phase, each node measures the number of neighbor nodes and determine whether it can join the cluster. This preclustering phase should be performed after initial network formation, and this phase should be performed whenever there is a change of the number of live nodes across network lifetime. In set-up phase, base station selects proper cluster head by measuring the amount of residual energy. Cluster head and non-cluster head report their residual energy to base station at the end of pre-clustering phase. Each elected cluster head broadcasts its information of being selected as cluster head with location information and cluster head identifier. Each non-cluster head can join the nearest cluster head. To complete cluster formation, each cluster head make a TDMA-schedule, and broadcast TDMA-schedule to cluster members. In steady-state phase, each node can send data to designated cluster head during allocated time slot. Cluster head aggregates date from cluster member nodes to send to sink node. After determining the set of cluster heads for next round, BS notify the result to cluster heads. Each cluster head i broadcasts ADV_CH(i) message to all nodes to notify node i can be candidate of cluster head for noncluster head j in the field. Each non-cluster head selects a proper cluster head based on their location. Then, non-cluster head i sends Join_Cluster(i, j) message to cluster head j to indicate joining to cluster j. After completing cluster formation, each cluster head makes TDMA schedule for non-cluster heads, and sends the schedule to the nodes in the cluster. Cluster head selection based on residual energy is illustrated in Figure 6. Initially, base station elects cluster heads on basis of the amount of current residual energy in descending order for K clusters, shown in

113


equation (2). As time goes by, there will be change of the number of live nodes because some nodes will die due to total exhaustion of energy. A new value of K’ will be recalculated.

4.2 Calculation of the Amount of Energy Consumption To evaluate performance of the HD-LEACH, we assume the same energy consumption mode1 as that of both LEACH and D-LEACH, where the radio dissipates Eelec = 50 nJ/bit to run the transmitter or receiver circuitry and εamp = 100 pJ/bit/m2 for the transmit amplifier in Figure 7. We make the assumption that the radio channel is symmetric so that the energy required to transmit a message from node A to node B is the same as the energy required to transmit a message from node B to node A for a given SNR. For our experiments, we also assume that all sensors are sensing the environment at a fixed rate and thus always have data to send to the end-user.

Figure 7. Energy consumption model

Figure 5. Operational procedure of HD-LEACH manner

non_ CH(i,r)

non_ CH(i,1) CHi

(1)

non_ CH(s,1)

non_ CH(j,r)

ing E(precluster round )

CHs (1)

(3) Select K CHs for next round (r+1)

(1)

non_ CH(j,1)

non_ CH(s,r)

BS

(1) Notify residual energy Of each node after round r (2) Sort residual energy of all nodes in descending order.

non_ CH(k,r)

non_ CH(k,1)

CHj

CHk

Figure 6. Base station gathers current amount of residual energy of both cluster heads (CH) and non-cluster heads (non_CH), and determines proper CHs for next round (r+1)

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Total amount of energy dissipation in HDLEACH may be computed similarly as LEACH except for pre-clustering phase and setup phase. The amount of energy dissipation during preclustering phase which decides joining nodes for this round may be expressed as (l disco.  Eelec  l disco.   ampR 2 )     (2)  N   N   disco.   Eelec  Elocal _ density    k   1  l     

where round  1, k  N, k is the suggested number of clusters, N is the number of initial live nodes. ldisco. is the size of Discovery-packet from a node i to neighbor nodes in radius of R. Each node i compute Di of itself by counting the number of messages received from neighbor nodes of node i.

114


The amount of energy dissipation of both cluster heads and non-cluster heads in setup phase may respectively be described as follows. _ for _ nonCH _ for _ nonCH E(setup  E(Decision  round,i , j ) round,i )

Number_ of _ CHs CH _ Advertise r r 1

l

 Eelec

preclustering steady _ for _ cluster E(total  E(setup round,i , j )  E( round ) round,i , j )  E( i , j )

5 PERFORMANCE EVALUATION

_ for _ CH _ of _ CH 2 E(setup  E(Selection  l CH _ Advertise   amp  d MAX  round,i , j ) round, j )

l

(4)

2  Eelec  ETDMA _ Schedule  l TDMA _ Schedule   amp  d MAX

Thus, we can write the amount of energy dissipation required at setup phase of a round in cluster j as follows. setup _ for _ nonCH _ for _ CH E(setup  E(setup round,i , j )  E( round,i , j ) round,i , j )

_ for _ cluster E(steady i, j )

,

The

amount

of

(9)

(3)

2  ESelect _ CH  l Join   amp  d nonCH  l TDMA _ Schedule  Eelec i , CH j

Number_ of _ Join _ Messages Join _ Message i i 1

Consequently, total energy dissipation per a round in HD-LEACH may be expressed as follows.

(5) energy

dissipation in steady-state stage, which considers operations of both non-CHs and CH in a cluster j may be expressed as

5.1 Computer Simulation Environment Simulation parameters and values are shown in Table 2. We assume that BS are deployed on coordinate (50, 50), which is in the center of field. Each node consumes energy of Eelec=50nJ/bit to process data in the node, εamp=100pJ/bit/m2 to send data to other node. And each cluster head consumes energy of Egather=5nJ/bit/message to gather data from joined non-cluster heads in its own cluster. We enumerate the details of simulation parameters in Table 2. Table 2. Simulation parameters

steady _ for _ nonCH ( l ,i , j )

E

 l  Eelec  l   amp  d

2 nonCHi ,CH j

(6)

Parameters Area for WSN

  N "   N " _ for _ CH E(steady      1  l  Eelec     l  Edata _ aggr. l ,l ', j ) (7)  k'    k'   4  Eelec  l ' amp  d CH j , SINK   N "  _ for _ cluster _ for _ nonCH _ for _ CH E(steady      1  E(steady  E(steady i, j ) l ,i , j ) l, j) k '     k'

 (2 N "k ' )  l  Eelec  N "l  Edata _ aggr.  l 'Eelec   amp   d j 1

(8)

4 CH j , SINK

 N"  1 k '  k ' 

 ( N "k ' )  l   amp   j 1

d i 1

2 nonCHi ,CH j

where 1  i, j, k’  N” and k’ is the number of clusters in current round, N” is the number of joining or active nodes after pre-clustering phase in a round and l’, l’  l is the size of aggregated data for transmitting from a cluster head j to sink node.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

The number of nodes

Value 10,000

m2

(100 m  100 m) 100

The location of sink node

(50, 350)

The location of base station

(50, 50)

Eelec

50 nJ/bit

Eamp

100 nJ/bit

Egather

5 nJ/bit

Packet Size

500 bytes

Packet size by Base Station

26 bytes

Initial residual energy

1 Joule

P: the ratio of CHs to all nodes

5%

Sensor node distribution model

Uniform distribution, Changes of node density from 40% to 80% by 20%

Software architecture of implemented simulator is illustrated in Figure 8.

115


Start

Initiation()

Coordinate()

Performance Metrics Analysis

Pre_clustering()

Set_up()

Steady_state()

Network Distribution Model

discrepancy of both cluster size and the number of clusters in a network. Also, it implies that HD-LEACH can reduce total amount of energy consumption in a round compare to LEACH and D-LEACH, as a result prolong the network lifetime of WSN.

Local Node Density & CH selection schemes

Energy Consumption Model

End

Figure 8. Software Architecture of Computer Simulator

It is crucial to control the density of node for comparing the network lifespan and the energy dissipation in three different clustering schemes including LEACH, D-LEACH and HD-LEACH. In computer simulation, we provide various scenarios for node distribution, where certain percentages of nodes among all nodes are intentionally distributed in a quadrant area over entire area of field. Thus, some nodes are placed in specific portion of entire field for performance evaluation with change of node density. 5.2 Performance Metrics and Evaluations

Figure 9. Example of cluster shape in HD-LEACH

5.2.2 The Number of Clusters We compare the occurrence ratio of number of clusters of each clustering scheme including LEACH, D-LEACH and proposed HD-LEACH by computer simulation. In figure 10, there is no great difference between LEACH and DLEACH according to varying the number of live nodes. However, there is great difference between HD-LEACH and others in terms of the average number of cluster. Especially, HDLEACH maintains five clusters with about occurrence ratio of 95 percent across total rounds of life of sensor nodes. This means that there is relatively low disparity of the number of clusters in HD-LEACH.

5.2.1 The Shape of Clusters Figure 9 shows an example of the cluster formed with HD-LEACH. The number of clusters is 5 and all clusters are balanced in terms of cluster size. All clusters in HDLEACH have similar shape with lower standard deviation of cluster size. There is a little difference of cluster size. These properties are different from LEACH and D-LEACH. This means that balanced shape of cluster and similarity of cluster size can reduce the

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 10. Occurrence ratio of the number of clusters among three clustering schemes across total rounds

116


5.2.3 The Cluster Size We compare the occurrence ratio according to cluster size among LEACH, D-LEACH and proposed HD-LEACH by performing computer simulation with parameters in Table 2. In Figure 11, there is a little difference in view of the average cluster size among three clustering schemes according to varying the number of live nodes. However, standard deviation of cluster size HD-LEACH is relatively low in comparison to LEACH and D-LEACH. In case of hybrid D-LEACH, this indicates that each cluster has relatively evenly joined nodes in view of the number of non-cluster heads in each cluster. This implies that all nodes are similar in amount of residual energy across whole network lifetime. Consequently, HD-LEACH can provide balancing of cluster sizes and residual energy of each node, resulting in prolonged network life.

Figure 11. Comparison of occurrence ratio of cluster size with the change of cluster size among three clustering schemes

We also show simulation results on cluster properties by HD-LEACH in Table 3. Comparing with cluster properties of existing schemes in Table 1, both average number of clusters and average cluster size by HDLEACH are similar to those of existing two schemes. However, HD-LEACH reduces standard deviation of average number of

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

clusters and standard deviation of cluster size. These imply that HD-LEACH generates more balanced clusters, which leads to potential increase of network lifetime. Table 3. Properties of clusters formed by HD-LEACH Number of live nodes Properties

100 (A)

65~99 (B)

30~65 (C)

0~29 (D)

Avg. number of clusters Std. dev. of number of clusters Avg. Cluster Size Std. Dev. of Cluster Size

5

4.3

2.3

1.1

0

0.8

0.4

19 8.9

19.6 9

20.4 7.5

0.2 10.6 6.8

5.2.4 Network Lifetime We can show the extension of network life time by applying HD-LEACH in Figure 12. To measure the effectiveness of network lifetime, we performed computer simulation with four different local node densities including uniform distribution. As local node density is higher, the difference of extended network life between HD-LEACH and two existing schemes is greater. In both LEACH and D-LEACH, as the local density increases, total amount of energy dissipation in a network suddenly increases due to energy dissipation rises of cluster heads after a node firstly dies. Thus, network lifespan can be reduced. However, network lifespan in HDLEACH can remarkably increase in comparing with both LEACH and D-LEACH. So, the round time that the first node dies is greatly delayed comparing with those of the others respectively as shown in Fig. 8. This means that HD-LEACH extends stable networking environment efficiently. In HD-LEACH, as the number of nodes deployed in area with higher node density increase, each live node consumes smaller amount of energy, which results in prolonged network lifetime. Consequently, HD-LEACH is superior to both LEACH and D-LEACH in view of extending of network lifetime due to balanced cluster size.

117


(a) In case all nodes are uniformly distributed over entire area

(d) In case 80 nodes are deployed in a quadrant, and 20 nodes are in the other three quadrants Figure 12. The progress of the number of live nodes with the change of local node density

6 CONCLUSION

(b) In case 40 nodes are deployed in a quadrant, and 60 nodes are in the other three quadrants

In this paper, we described HD-LEACH with a centralized clustering method that balances energy usage of each node by properly selecting cluster heads based on amount of residual energy. Different from existing LEACH and DLEACH, base station, selects proper cluster heads by considering residual energy of each sensor node at each setup phase in HD-LEACH. Our simulations show that HD-LEACH reduces the disparity of the number of clusters and cluster size at each round. Simulation results show that HD-LEACH greatly extends the network lifetime by generating more balanced clusters. 7 REFERENCES [1]

[2]

[3]

(c) In case 60 nodes are deployed in a quadrant, and 40 nodes are in the other three quadrants

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-Efficient Communication Protocol for Wireless Microsensor Networks”, Proceedings of the 33rd HICSS 2000 , IEEE Press, Volume 8, Issue 2, pp. 8020-8029, 2000 S. Banerjee, and S. Khuller, “A Clustering Scheme for Hierarchical Control in Multi-hop Wireless Networks”, Proceedings of IEEE INFOCOM, IEEE Press, Volume 2, pp. 1028-1037, 2001 M. Chatterjee, S. K. Das, and D. Turgut, “WCA: A Weighted Clustering Algorithm for Mobile Ad Hoc Networks”, Cluster Computing, Volume 5, pp 193– 204, 2002

118


[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

S. Bandyopadhyay, and E. Coyle, “An EnergyEfficient Hierarchical Clustering Algorithm for Wireless Sensor Networks”, Proceedings of IEEE INFOCOM, IEEE Press, Volume 3, pp. 1713-1723, 2003 O. Younis, and S. Fahmy, “HEED: A Hybrid, Energy-Efficient, Distributed Clustering Approach for Ad Hoc Sensor Networks”, IEEE Trans. Mobile Computing, Volume 3, Number 4, pp. 366-379, 2004. J. Kim, and T. Byun, “A Density-based Clustering Scheme for Wireless Sensor Networks”, Communications in Computer and Information Science, Volume 195, pp. 267–276, 2011 J. Kim, and T. Byun, “A Performance of a Novel Clustering Scheme Considering Local Node Density over WSN”, Communications in Computer and Information Science, Volume 266, pp. 320–329, 2011 J. Oh, S. Jang, and T. Byun, “A Centralized Cluster Head Selection Scheme for Reducing Discrepancy among Clusters over WSN”, Lecture Notes in Electrical Engineering, Volume 181, pp. 699–706, 2012 S. Ehsan, and B. Hamdaoui, “A Survey on EnergyEfficient Routing Technigues with QoS Assurances for Wireless Multimedia Sensor Networks”, IEEE Communications Survey & Tutorials, Volume 14, Issue 2, pp. 265-278, 2012 H. Park, H. Ahn, K. Park, and D. Kim, “EnergyEfficient Sensor Node Algorithm to Prolong Sensor Networks”, Lecture Notes in Electrical Engineering, Volume 181, pp. 423–427, 2012 N. A. Pantazis, S. A. Nikolidakis, and D. D. Vergados, “Energy-Efficient Routing Protocols in Wireless Sensor Networks: A Survey”, IEEE Communications Survey & Tutorials, Volume 15, Issue 2, pp. 551-591, 2013 B. B. Krishna, and A. S. Raghuvanshi, "Centralized border node based cluster balancing protocol for wireless sensor networks", Proceeding of 6th International Conference on Contemporary Computing, pp. 35 – 40, 2013 Z. U. A. Jaffri, and Cai Yueping "ZET: Zone and energy threshold based clustering routing protocol for Wireless Sensor Networks", Proceedings of 23th International Conference on Computer Communication and Networks (ICCCN), pp. 1–6, 2014

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

119


Connectivity Issues on IoT Business - The Korean Case of Smart Home Network Ji Yeon Cho1, Hye Sun Lee2, Bong Gyou Lee1* Graduate School of Information, Yonsei University 2 Technology and Business Administration, Yonsei University [email protected], [email protected], [email protected]* 1

ABSTRACT Connectivity is one of the most important factor that may affect as a great benefit to various domains for the next era of the IoT(Internet of Things). However, securing connectivity requires various industry stakeholders to access freely which cause business and regulatory issues. The purpose of this study is to explore challenges in implementing market based IoT industry, focusing on connectivity issues. We have investigated cases of the smart home industry in Korea, presented implications based on the literatures and the in-depth interviews. As a result, newly emerging industry based on technology like IoT tend to have conflicts between service providers and network providers. And it expands towards political issues on data traffic and fostering spectrum managements by private sectors. Moreover, connectivity with device-platform-service protocols and compatibility tend to be monopolistic structured. Our implications suggested supplementing ideas for improving ways of establishing regulations. KEYWORDS IoT(Internet of Things), Smart home, Connectivity, IoT Business, Spectrum policy

1 INTRODUCTION IoT technology has more advanced abilities and social impact compare to smartphones. In other words, IoT is an intelligent service from collect and generated information, analyze and share them by the Artificial Intelligence [1, 2].

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

It is obvious that the expansion of connectivity will bring the big change beyond smartphone. Global IT companies – Google, Cisco and etc have launched IoT related products and services. Korea government has been preparing for IoT industry as a new growth engine [3]. Major Korean enterprises have also begun to make inroads into the smart home markets [2]. Currently, the market players are from wide ranges of industry sectors - positioned as a different domain of the value chain, launching smart home services with the purpose of sustaining their market dominance. Thus, this can work as entry barrier for others introducing IoT services. The tendency as explained above directly effects on service connectivity – full scaled IoT environment, will eventually change the value chain structure [4]. Moreover, IoT connectivity may raise several key issues similarly with the past experience of smart device industry such as IPTV and Smart TV. Korea government has struggled with network occupancy subject issues and management with IPTV stakeholders and Mobile industry regarding network neutrality, interconnection [5]. Therefore, our research question is “what challenges will be emerging to support IoT industries competitiveness?” This study is organized as follows. In chapter 2, importance of IoT connectivity and smart home network were reviewed. In chapter 3, we have analyzed smart home industry in Korea. Finally comprehensive implications from both industrial and political perspective regarding IoT connectivity were given in chapter 4.

120


2 RESEARCH BACKGROUND 2.1 The Effects of IoT According to the Gartner, the number of connected devices in the Internet of Things will increase 30 times, approximately 26 billion devices by 2020. When Kevin Ashton first used term ‘IoT’, definition is focusing on utilization of RFID and sensors throughout the daily activities. Since then, the notion of IoT has been constantly changing according to the development of technologies. Nowadays, IoT means advanced connectivity of anything such as devices, systems, and service and people at any time in anyplace [1, 2, 3]. As growing the number of connected device and related industry, IoT is one of the most hyped technology trends in recent years. IoT technologies have significant impact on not only ICT business but also other converged industry. Jeong et al.(2013) examined economic effects of IoT industry in Korea through the Input-output Analysis [6]. As a result, the economic effect of IoT industry on production inducement is over 400 million dollars; effect on value-added inducement is over 300 million dollars during 2013 and 2017 in Korea domestic industry. According to the GSMA(2012), connected devices will make revenue in a global business as much as 4.5 trillion dollars by 2020 through facilitating new business models, and improving existing services [7]. This impact is caused by extension of the connectivity with everything. This is the reason why IoT connectivity issues have been studied in different domains as common research interest [4]. As illustrated in Figure.1, IoT connectivity issues have been discussed a lot with various perspectives such as devices, sensors, infrastructures and services [8]. 2.2 IoT Connectivity IoT connectivity issues are studied a lot in technical perspective including the evaluating network technologies, testing IoT device and

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 1. The perspective of IoT Connectivity modified from M. Skilton. [8]

studies related to IoT platform. In the connected environment, it is very important to make any devices connect all the time with the most efficient connection methods. Accordingly, IoT connectivity technology studies are tend to focus on finding way to interconnect any product in the physical world with the virtual world through the any network [9]. So this issue has been studied a lot in interconnected real network systems regarding finding methods to improving network efficiency and related wireless sensor networks [10]. Also connectivity issue can be found in existing studies regarding analyzing IoT architecture and technologies. According to M. Abomhara et al.(2014), IoT supporting technologies concept are mainly categorized as following: 1) identification technologies 2) networks and communication technologies 3) software and hardware technologies [9]. Architecture is basically consists of 4 layers including contents layer, service layers, network layers, device layers [11]. IoT has a more complicated architecture because of increasing connected device and its different range of networks. M2M architecture is almost similar to IoT. Including various range network protocol and platform [4]. And this architecture identified three main IoT technical domains which are device, connectivity and application

121


Figure 2. Roles in an IoT ecosystem [13]

service domain. Studies on IoT architecture have something in common that connectivity domain were emphasized. The key architecture components of IoT are focusing to support all connected environment. Research approach on the IoT connectivity according to this architecture tends to focus on technologycentric issues. However, nowadays IoT agenda has interests on not only the technology augmentation but also creating an additional value for other industry [2]. In this context, IoT connectivity studies with social science perspective are needed at the moment. The ecological approach is useful to understand comprehensive industry structure and to find out critical issues from interaction among participants [12]. Several studies on IoT value chain and ecosystem have been made and it is focus on deriving IoT business boundary and analyzing business model. Compare to studies regarding ICT convergence industry value chain which are identifying industry based on C-P-N-D(Contents, Platform, Network, Device) concept, IoT ecosystem studies identified various entity and roles of player related to connectivity. This approach to IoT industry will be improve to understand issues regarding political issues, value chain and business model in the various industry level. According to Mazhelis, O. et al.(2012)[13], core of IoT ecosystem should be explained interconnection

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

of things. Figure 2 is the map of IoT ecosystem and key players [13]. The key player’s roles in the figure 3 are mainly organized as followings: 1) service delivery dimension, 2) connectivity 3) device. The roles of player are also categorized with the concept of lifecycle which is consisting of development, distribution, provisioning, assurance, and billing. One of the critical point of this ecosystem framework is that they are explains the importance of the political perspective. The legislative, regulatory, standardization roles are regarded as essential roles on IoT ecosystem in this study. Korean learned from the experiences in smart phone industry as well as IPTV which is convergence of broadcasting and telecommunication industry. Because of the technological advancements, the boundaries between broadcasting and telecommunications are confused. This industrial change makes it difficult to handle the conflict with the current regulatory. The current regulation framework cannot be applied on the new connected environment with IoT technology. It is expected to cause players in the IoT industry feel as if they are being subjected to an inconsistent regulatory system. Consequently, it is very important to tackle the relevant issues of a regulatory framework from the beginning of making business ecosystem.

122


Figure 3. Various industries related to smart home (Home IoT) [17]

2.3 Smart Home with IoT The IoT Technologies can be converged in a variety of industry and Smart home is considered as one of the most promising IoT application domains [3].Naturally, smart home became the center of interest for both scholars and global leading companies. Google has acquired Nest Labs, smart home appliances start-up, for 3.2 billion dollar in 2014. Samsung electronics also announced the plan to achieve a leading position in emerging businesses such as smart home industry on IoT platform. They continuously released the innovative home appliance at the CES tradeshow and have entered into IoT platform business for smart home [14]. The Smart home concept has been studied for the last decades with different name such as home network, ubiquitous home and digital home [4]. Smart home can be defined as a residence that use an advanced ICT technology to promote their convenience, security, entertainment and communication through the home network within the home and connections to the world [15, 16]. In other words, Smart home is a phenomenon that ICT technology is applied to daily life at home and related area. The range of smart home field is expanding rapidly as IoT technologies are converged [16]. Many researches on smart home defined smart home and identified related business with broad perspective. As shown in Figure 3, smart home encompasses a variety of

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

industry and applications including telecommunications, home appliance, education, security and healthcare etc [17]. To support this various application service under connected environment, wired and wireless home networking technologies considered as core and basic domain in the smart home era. Because smart home environment is subject to a connected everything in an intelligent way. KCA(Korea Consumer Agency) also place a stress upon ubiquitous home network system which is convenient to manage home environment. According to KCA, stable wired and wireless network system must be established first to make smart living space. Common points in recent definitions of smart home are as follows: 1) electric home appliances and service are getting smart. 2) IoT function and concept is applied 3) network environment supports IoT business 4) devices are making information and exchange them itself. 5) Collected information makes various services and identifies consumer demand. Although smart home industry is not matured enough to promote innovative IoT application service, it is expected to market growth through diffusion of the communication function equipped device. Figure 4 shows the changes of connected device market structure according to communication module [18]. When completely new product with connectivity is emerging in the smart home industry, the largest new market creation effect will be expected.

123


Figure 4. Connected device market structure: Traditional vs. New market [18] Table 1 Main Components of the Smart Home Ecosystem [17] ECOSYSTEM SUB ELEMENTS ELEMENTS Wired network VDSL, FTTH, GiGA Network Wireless network LTE, nationwide broadband LTE-Advanced Home appliance White goods Smart Device New smart device Energy, security, health related device All smart devices IoT Standards B2B Business area Home Hub(device/cloud) HomeKit, Nest, TV, Set-top etc. Platform Operation O/S iOS, Android, Tizen Control device TV, PC, smart phone, tablet Display Home appliance contents Home appliance remote control, Intelligent control Content New smart device content Energy, security, health related service

This is similar to the phenomenon in the mobile ecosystem which is mobile technology get applied in the various industries. As previously mentioned, study with ecological perspective is useful to identify the industrial issues and to find business strategy. Table 1 is ecosystem framework for smart home under the IoT environment that it is difference in including IoT standards and platform. Device and content level in C-P-N-D value chain was divided in more detail. Device level is specified into home appliances and smart device and content level is also identified new type of content service. This changed ecosystem has effects on key player’s business strategy regarding smart home device development, network connection methods and service diversification based on generated data from connected things. All these issues are directly related to connectivity issues and collaboration with other key players same as in ISBN: 978-1-9491968-07-9 ©2015 SDIWC

smart phone ecosystem. For this reason, smart home industry in IoT paradigm is expecting to be a battle ground as secondary mobile warfare [19]. The strategic movements among the key operators from various fields are expected to secure a leading position in the smart home ecosystem with their competiveness. 3 CASE ANALYSIS OF KOREA 3.1 Research Methodology We have analyzed Korea smart home network industry in order to draw issues on smart home connectivity with IoT convergence. Because smart home and IoT embodiment in Korea is rapidly actualizing yet it is still ongoing activities that can be seen as a contemporary phenomenon [20].

124


Table 3. Visible Presence of the Smart Home Market

Table 2. Interviewed Experts Domain EXPERT’S DOMAIN

POSITION

Smart Home(including Home Network), IoT, M2M, Smart Plug and Smart Grid service, Building Constructor, Device/Appliance manufacturer

Professors, CEOs, Managers, Executive director, Consultants

Note: Interviewed 10 experts (+10 years of work experience)

Thus, we have conducted in-depth interviews as a data gathering method with smart home and IoT related industry experts from August 18th to 27th in 2014. As a result, total 10 experts with work experiences of over 10 years in this domain have been selected for face-to-face indepth interviews. 3.2 The Market Movements of the Smart Home Industry in Korea Korean Smart Home industry is rapidly converging with the IoT sector by introducing products that are compatible with smart home services. In this study, we have categorized into two groups that can be described as visible presence of the current market movements – home appliances and newly commercializing products(See Table 3) based on in-depth interview. 3.2.1 Home appliances market Home appliances market in Korea consists with various local firms but the market size of smart home products has been majorly led by Samsung and LG for the past few years. Overall market trend including minor products(eg. rice cooker and etc.) are tend to introduce various features of smart control with the purpose of occupying market dominance. Firstly, for the case of general home appliance manufacturers are introducing smart products that connect manufacturer’s platform by Wi-Fi

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

HOME APPLIANCES Digital TV, Refrigerator, Washing machine, Vacuum cleaner washer, Microwave

NEWLY COMMERCIALIZING Black box, Smart Meters, Digital sensors & Alarms,

Secondly, Samsung electronics’ home appliances are embedding smart applications features for the new products and supplying an external device(dongle) for legacy system [21]. Currently, Samsung SDS, which is a subsidiary of Samsung group providing information technology services, is staring business on smart homenet application in android and iOS connecting with their wallpad – existing home network system installed in residents or Raemian apartments(Samsung C&T)[22]. Lastly, LG is introducing technology named by ThinQ and homechat, focusing on realizing Artificial Intelligence service for automated home appliance voice control. The aim of this technology is to support their appliances to be controlled over the smartphones as well as automated diagnosis and upgrade firmware, save electricity, smart management by Wi-Fi and ThinQ. The two companies are based on their home appliances communicating with smartphones but soon will shift into smart TV for an overall control box. Thus, characteristics of the home appliance sector can be described as follows.  Home appliances manufactures are expanding inherent industry sector towards own service platform based on the strength of product distribution.  Samsung and LG are planning to expand full-scale of the smart home infrastructure for the coming IoT service dominance.  Smart Home application and different brand of appliances are not compatible.

125


Figure 5. Visible presence of the Smart Home Players Modified from [17, 23]

3.2.2 Newly commercializing product market Newly commercializing products that we have categorized are sensors and devices, can be connect with the smart home environment CCTV, energy saving and monitoring for lighting, thermostats, home security, u-health, home hub and etc. However, currently there are no distinctively dominant products or service providers in the market. That is these categorized products and services may be individual yet relates with large number of stakeholders across the industries. Therefore, it has been a time consuming progress for the standardization or unification to collaborate with another for providing services. Current market movements within this sector tend to be supporting more than one standard formats rather than unifying service formats by single consortium. Although the market is in initial stage, there are visible presences of potential players mutually competing for the platform dominance - telecommunication operators, IT firms, home appliance vendors(See Figure 5). Firstly, Korean telecommunication operators are focusing on dominating platform. For the case of SKTelecom, has recently announced business plan smart home. They are targeting to compete with Samsung and LG in home appliance sector by collaborating with other local home appliances firms(See Table 4). SKT has picked up problems of mutual compatibility between manufacturers and products [24, 25]. Thus, SKT is aiming to release various smart

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

home products with high market share and provide connectivity by their wired and wireless services. However, as a telecommunication company, SKT has to charge service rates to manufacturers or users that there was an announcement for plan yet no specific rate is opened. On the other hand, the case of LG telecom is collaborating with affiliated companies. Unlike Samsung, LG has a telecom company that makes their strategy different. Samsung is focusing on developing standards while LG is planning to collaborate with others to fulfill better features of LG’s consumer products. Table 4. Consortium of SKT and the Local Firms [25] PRODUCTS MANUFACTURERS Boiler/ Heater Door Lock

KDNavien

1st market share of boiler

DSceltic

Condensing boiler

Irevo CLK corp

Home Appliances

in

the Air

Winix corp

Dehumidifier

Monueual

Robot vacuum cleaner

Yujin Robot

Robot vacuum cleaner

GE Lighting Lighting

1st position market Carrier, conditioning

Kumho Electric

AP

IPTime

Gas Circuit breaker

Time Valve

Global lighting company Local lighting company 1st position in Wi-Fi AP Timer type gas circuit breaker

126


That is LG doesn’t have intention to develop own smart hub device to avoid conflict with other telecom operators [26]. But in the same time, LG U+ adopted Z-Wave for the next home network protocol differentiate with others. SKT is planning to shift LTE smartphone users to SKT smart home network rather than enabling Zigbee or Bluetooth [27]. Therefore, presently telecom companies are in the initial progress to see the market consumptions and they are preparing for brief business model. Thus, various wireless protocols cannot be coexist in the market due to customer’s cost benefit. Universal connection across the products and networks should be waited until the dominant de facto protocol appears. Secondly, IT platform in smart home service act as a passage between devices and contents. Like other countries, Korean OS platform has not been widely used as Apple or Google’s. With the smart home service platform, currently Samsung and LG are looking into TV for embed and diffuse as own platform(see Table 5). Samsung has embedded Tizen 3.0 in smartTV that can be compatible with android OS and contents for wearable devices. LG is planning to use smartphones or mobile devices as the platform rather than TV. Field experts are seeing this strategy from status of webOS that it needs more time compare to Tizen [28]. On the other hand, there are ongoing progress by the ones who has been owned market dominance in traditional industries such as KDN(Korea Electric Power Data Network Co., Ltd.) for smart grid, major security companies and etc. These firms’ approach towards smart home is very similar with telecom companies, Table 5. Samsung and LG’s platform SAMSUNG

LG

Control method

Voice

Text

Communication standards

OIC

AllJoyn

TV OS

Tizen 3.0

Web OS

Smart home Hub

M&A Smart things

Collaborate with Telecom company

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

but the service network and business model is very independent. Thus, characteristics of the newly commercializing product market sector can be described as follows.  Players in this area are from wide range of industry sectors - manufacturers, technology developers, telecommunication providers, service providers and energy suppliers.  The market is currently in initial progress that there is no visible dominant product or service.  Most of the commercialized products(final products yet close to test the market) are based on Wi-Fi protocols. 4 IMPLICATIONS 4.1 Industrial Implication for Connectivity The framework was drawn through the literature review and experts interview that reflect key player’s originality and their market positioning [4, 17, 18, 30]. Firstly, the most critical implication in the industrial perspective is identifying the stakeholder’s connectivity position. In order to make connectivity more substantial, some experts have argued that stakeholders should understand their diversity and make an effort to define exact market position in smart home market. In other words, connectivity issues begin from market positioning and that is directly related to connectivity solutions. We divided smart home market into two groups based on the device characteristics. First part is home appliance market which is traditionally dominated by global leading company such as Samsung electronics and LG electronics. Second part is newly commercializing product market involved many companies in making new type of smart home service. This category is related to original competiveness and to roles of player in smart home ecosystem. If business operators defined their market position considering competiveness, device life cycle and service level, then they should try to find out connectivity position.

127


Figure 6. Conceptual Framework for IoT Connectivity Strategy

Competition among key players around connectivity can be occurred every connectivity level regarding communication network, platform, device standard. For Korea smart home ecosystem, some experts emphasized that the basic connectivity with wired and wireless internet access will be the first and the foremost in the beginning stage of smart home ecosystem with IoT. Connectivity related to devices standard and platform is following after. Past experience in mobile ecosystem shows that when data traffic increasing rapidly, network operators has lost their revenue base and uncertainty of return on investment in network infrastructure is expected. It has significant impact on destroying virtuous cycle of mobile ecosystem and this phenomenon eventually causes the frequency issues and the decline competitiveness of connectivity [29]. Therefore, to make a successful connectivity strategy in smart home era, key players should understand their connectivity position and keep favorable relationship with network operator. Second important industrial implication is related to providing smart home service level. According to the evolving smart home service level, connectivity issues regarding capabilities of supporting network, data exchange and industry boundaries are different. M. Porter et al.(2014) has explained about transforming competition with the capabilities

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

of smart connected products which are four areas as shown in figure 6: monitoring, control, optimization, and autonomy [30]. The evolution of service in smart home means encouraging innovative business model and redefining industry boundaries through connecting everything based on connectivity. The relationship between service operators will be very different depends on the service level positioning. For example, if entrepreneur aims at doing business within his origin industries, connectivity capability with other partner is may not the main concern. However, if he plans to expand industry boundaries and to have leading position in convergence industry by providing comprehensive service, it is expected that integrated systems and collaboration strategies will be the essential. To sum up, through this conceptual framework, regulatory issues in connectivity can be identified. In all connected industry, stakeholders from different industry can be located in same connectivity position. So, setting the balanced regulation system is crucial with understanding of connectivity issues in business perspective. 4.2 Political Implications on Spectrum Issues For the next IoT era, connectivity of the smart home network needs to be accessed freely for

128


users and the stakeholders in the value chain. Currently, Korea is using a very large number of wireless communication devices for 2.4GHz band - ISM for wireless LAN such as wireless home phones, Bluetooth devices. This means when IoT with personal preference services are provided as a full-scale, it is expected shortage of frequency bandwidth [31]. Currently, most of the Korean smart home services and products are based on Wi-Fi protocols rather than low powered protocols. Therefore, scarcity of the spectrum resources needs to take complementary measures to promote the smart home industry. One of the complementary measures that may supplement the current spectrum issues can be an open access. Based on the expert interview, we draw political implications on realizing open access as 1) deregulation on unlicensed band and spectrum sharing and 2) considerations of market based spectrum management. Firstly, value of utilizing unlicensed band can be very high, estimated within U.S.A is assumed to be over 140 billion US dollars by annually [32]. Unlicensed band may meet the requirements of smart home services - a low powered and flexibility for universal connectivity. Korean government is also actively looking forward to support the spectrum sharing technology and DSA(Dynamic Spectrum Access) for the industrial competiveness. However, government’s execution on allocating new band or deregulation often challenges practical issues as well as time consuming for considering stakeholders position. Thus, it can be a key that government promptly produces additional measure to redress all the related possible drawbacks of the spectrum policy to go along with the market changes. Secondly, commercial band manager and secondary spectrum market needs to be examined thoroughly in short period time for ensuring more efficient spectrum usage [33]. This is because realistically it is difficult to secure a new frequency bands that have mostly assigned below 3GHz(suitable for mobile

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

communication). Also, 700MHz bandwidth that is digital dividend by digitization of terrestrial broadcasting is having conflicts for its usage with various sectorial stakeholders’ demands. 5 CONCLUSION This paper has explored a smart home market of Korea to assess barriers in implementing market based IoT industry, focusing on connectivity issues. Full-scales of IoT service demands by interconnecting various devices and information may raise new policy issues. In the case of new technology based industry where in formative period, it is important for the government to policy making and settle potential conflicts of business players. From the past experience of IPTV and smart TV, Korean government has been investigating various ways to promote flexible and innovative IoT ecosystem. In this paper, implications have drawn based on the literature reviews and in-depth interview to explore current status of the Korean smart home industry. As a result, some of the barriers found in the case analysis were political issues on fostering spectrum managements by private sectors. And connectivity with device-platformservice protocols and compatibility tend to be monopolistic structured. Thus, the established firms from the traditional industries are beginning to strategize to sustain privilege by expanding their infrastructure for occupying smart home serviced platform in advance. That is home appliance manufacturers are framing a service platform based on their devices, or telecom operator tend to frame a service platform affiliating with other smart devices based on their wired and wireless platform. Thus, smart devices with enhanced inter-device communication is important to the firms for leading smart systems with high degrees of intelligence and autonomy, facilitating the rapid IoT application deployment and creating new services [18]. The suggested barriers and implications assumed to supplement idea for improving ways of establishing regulations.

129


Blog,

Moreover, conditions for optimum allocation of industry positioning between private and public sectors of IoT will need to be examined for further study. [9]

ACKNOWLEDGEMENT "This research was supported by the MSIP(Ministry of Science, ICT and Future Planning), Korea, under the ITRC(Information Technology Research Center) support program (NIPA-2014-H0301-14-1042) supervised by the NIPA(National IT Industry Promotion Agency)"

[10]

[11]

REFERENCES [12] [1] [2]

[3] [4]

[5]

[6]

[7]

ITU, The Internet of Things, ITU-T Internet Reports, 2005 K. C. Park, J. J. Lee, S. H. Oh, and B. G. Lee, “Harmonizing government policies and enterprise strategies for IoT business", 3rd International Conference on Circuits, Systems, Communications, Computers and Applications (CSCCA'14), Florence, Italy, 2014.11 Ministry of Science, ICT and Future Planning, National Strategies for Internet of Things, 2014 O. Mazhelis, H. Warma, S. Leminen, P. Ahokangas, P. Pussinen, M. Rajahonka, R. Siuruainen, H. Okkonen, A. Shveykovskiy, and J. Myllykoski, “Internet-of-things market, value networks, and business models : State of the art report,” 2013. J. H. Kwak, J. Y. Cho, Y. S. Lee, and B. G. Lee,, “Remaking Mobile Ecosystem Policies for New Mobile Market”. Journal of Korean Society for Internet Information, Vol.12, No. 4, pp. 93-106, 2011 W. S. Jeong S. H. Kim, K. S. Min, “An Analysis of the Economic Effects for the IoT Industry”, Journal of Korean Society for Internet Information, Vol. 14, No. 5, pp.119-128. 2013. GSMA, “GSMA Announces the Business Impact of Connected Devices could be Worth US$4.5 Trillion in 2020”, Press relase Feburary 27, 2012, available online at http://www.gsma.com/newsroom/press-release/gsmaannounces-the-business-impact-of-connected-devices-could-beworth-us4-5-trillion-in-2020/, last retrived January 15,

[8]

2015. M. Skilton, “The era of “Internet aware systems and services” – the multiple-data, multi-platform and multi-device and sensors world”, The Open Group

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

available

online

at

http://blog.opengroup.org/2013/06/28/the-era-of-internet-awaresystems-and-services-the-multiple-data-multi-platform-and-multidevice-and-sensors-world/ , last retrieved January 15,

[13]

[14]

[15]

[16]

[17]

[18] [19]

2015. M. Abomhara and G. M. Koien. "Security and privacy in the Internet of Things: Current status and open issues." Privacy and Security in Mobile Systems (PRISMS), 2014 International Conference on. IEEE, 2014. B. S. Panda, D. P. Shetty, “Minimum Range Assignment Problem for Two Connectivity in Wireless Sensor Networks”, Distributed Computing and Internet Technology, Lecture Notes in Computer Science, Vol. 8337, pp 122-133, 2014. S. Turber, J.vom Brocke O. Gassmann, and E. Fleisch, “Designing Business Models in the Era of Internet of Things.”, Advancing the Impact of Design Science: Moving from Theory to Practice, pp. 17-31, Springer International Publishing, 2014 J.Y. Cho, T. Kim, K. C. Park, B. G.Lee, “An analysis of big data structure based on the ecological perspective”. Korea Soc. IT Serv, Vol. 11, No. 4, pp.277–294, 2012. O. Mazhelis, E. Luoma, and H. Warma, “Defining an internet-of-things ecosystem”, Internet of Things, Smart Spaces, and Next Generation Networking, pp. 1-14, Springer Berlin Heidelberg, 2012. S. Tibken, “Samsung, SmartThings and the open door to the smart home”, cnet CES 2015, January 12, 2015, available online at http://www.cnet.com/news/smartthings-ceo-on-samsung-beingopen-apples-homekit-and-more-q-a/ F. K. Aldrich, “Smart homes: past, present and future”, Inside the smart home, pp. 17-39, Springer London, 2003. R. J. Robles and T. H. Kim, “Applications, Systems and Methods in Smart Home Technology: A Review”, nternational Journal of Advanced Science & Technology, Vol. 15, 2010. Y.G. Kim. “Six maincomponents of the smart home ecosystem”, KT digieco report, Issue&Trend, November, 2014. J. H. Park, “IoT market: Truth and Falsehood”, KT digieco report, Issue&Trend, Augst, 2014. M. S. Kim, “Samsung began in full swing for smart home platform”, Korea Economic Daily, News relase November 13, 2014, available online at http://www.hankyung.com/news/app/newsview.php?aid=2014111 32065g

R. K. Yin, 2003. Case study research: design and methods, (3rd ed.). Thousand Oaks, Calif.: Sage Publications. [21] M. H. Yun. D. H. Jang, “Smart appliances Status and Future Direction”, KEIT PD Issue Report, Vol. 12. No.4,.pp. 145-163, 2012. [20]

130


[22]

Samsung

SDS

website,

available

online

at

http://www.sds.samsung.co.kr [23]

Trend Spectrum, “Samsung, Smart Home Service official launch”, April 02, 2014, available online at http://trendspectrum.co.kr/?p=16865

[24]

H. J. Kang, “Samsung , LG, ecosystem-based strategy”, Digital Times, October 07, 2014, available online at http://www.dt.co.kr/contents.html?article_no=2014100802101232 742001

[25]

H. J. Kang, “SK Telecom 'smart home Allies' launched”, eDaily, October 02, 2014, available online at http://www.edaily.co.kr/news/NewsRead.edy?SCD=JE31&newsi d=02368166606248984&DCD=A00503&OutLnkChk=Y

[26]

M. J. Ki, “[IFA 2014] Samsung Standard occupancy · LG products ... another way to differentiate Samsung · LG Smart Home”, The Korea Economic Daily, September 10, 2014, Available online at http://www.hankyung.com/news/app/newsview.php?aid=2014090 947151

E. J. Oh, “Carrier IoT standardization war started ... LGU +, write home network IoT communication standard `Z- wave`”, August 20, 2014, etnews, available online at http://www.oss.kr/oss_news/605656 [28] T. H. Yang, “Samsung vs LG web Tizen OS 'platform war winner”, December 19, 2014, inews, available online at http://www.oss.kr/oss_news/605656

[27]

[29] H. Y. Lim, “The evolution of ICT ecosystem and Business Strategy of Golobal ICT leader”, Information & Communications Magazine , The Korean Institute of Communications and Information Sciences, Vol. 28, No. 12, pp.16-23, 2011.

M. E. Porter and J. E. Heppelmann, “How smart, connected products are transforming competition”, Harvard Business Review. November. 2014. [31] J. T Oh, “A Study on the required Frequency Bandwidth for Personal Environment Service”, The journal of Korean Institute of Communications and Information Sciences, Vol. 34, No.11, 2009. [32] Columbia Institute for Tele-Information, "Assessment of the Economic Value of Unlicensed Specturm in the United Staes", 2014.2. [33] J. M. Peha, “Wireless communications and coexistence for smart environments”, Personal Communications, IEEE, Vol. 7, No. 5, pp. 66-68, 2000

[30]

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

131


Implementation of Digital Radio for Improved Synchronization using FPGA-based Platform HongSun An, Jaehong Kim, and Bongseop Song Neo Embedded System and Service Lab [email protected], [email protected], [email protected]

ABSTRACT In this work, a narrow band digital radio is implemented using the FPGA-based platform. We concentrate on the synchronization part which is located at front of digital modem. In peer to peer communications, synchronization between radios is very important. If the synchronization fails, the receiving radio cannot receive any signals. Therefore, the synchronization part plays a crucial role for digital radio system. We propose the effective synchronizer implementation method to improve the performance of digital radio receiver. The measurement results prove that the proposed synchronizer implementation scheme achieves the possible synchronization frequency range around eight times more than conventional synchronization system. Also, it is shown that the implemented digital radio achieves approximately -112 dBm received sensitivity at target FER of 10-2.

KEYWORDS Automatic Frequency Control, Synchronizer, Digital Radio, Frequency Offset, FPGA

1 INTRODUCTION The FCC (Federal Communications Commission) decided the date for the implementation of the 6.25 kHz equipment certification rules as January 1, 2011, but strongly urges licensees to consider migrating directly to 6.25 kHz radio technology by January 1, 2013 rather than initially adopting 12.5 kHz digital radio technology and later migrating to 6.25 kHz technology [1]. Since these decisions had been made, 6.25 kHz digital radio standards were finalized. For example, in Europe, this new

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

digital 6.25 kHz radio idea was taken up by ETSI (European Telecommunications Standards Institute) and developed into a European Standard called dPMR (digital Private Mobile Radio) [2], [3]. For the North American LMR (Land Mobile Radio) market in particular, the NXDN protocol was developed [4]. Moreover, the increase of spectrum demands in radio system using frequencies of VHF and UHF has required for the increase in the spectrum reuse efficiency. For alternative plans, changing from wide-band channel to narrow band channel and using multiple access technology which can increase 4~5 times of frequency reuse efficiency in comparison with current 25 kHz channel FM scheme have been reviewed. Considering these trends, the development of 6.25 kHz digital radio is becoming more and more important issue. In conventional communication systems, the synchronization is a very important technique. If any MS (mobile station) fails the synchronization, it cannot receive any signals from a BS (base station). The synchronization process is very sensitive to frequency offset, and frequency offset is generated by various factors, such as the mobility of MS, the clock offset of an oscillator, and the hardware thermal noise [5]. In a cellular system, all MSs attempt to synchronize based on a BS. However, in a radio system, BS does not exist. Due to this fact, the digital radio should perform the synchronization process efficiently even if the frequency offset still exists. We propose several implementation schemes for the narrow band digital radio modem. The frame structure and digital filter techniques are

132


proposed for supporting the high rate, and synchronizer implementation methods are suggested for the efficient synchronization process. This paper is organized as follows. Section 2 describes the frame structure and radio modem configurations of the implementtted digital radio. Section 3 presents the measurement environments and experimental results. Finally, section 4 draws the conclusions. 2 DIGITAL RADIO MODEM CONFIGURATION In this section, we first present the proposed frame structure for 8,000 bps 6.25 kHz digital radio. Then, we describe the used AMC (Adaptive Modulation and Coding) level and digital filter. Lastly, we focus on the synchronizer implementation scheme. 2.1 Frame Structure The proposed frame structure of digital radio for achieving maximum data rate 8,000 bps level using QPSK modulation is configured as the following figure.

Figure 1. The frame structure of proposed digital radio

The first 6.4ms part is the 32 preamble symbols, and it supports for the synchronization of receiver. The second 73.6ms part consists of the 46 pilot and 322 data symbols. The pilot symbols exist for wireless channel estimation and frequency offset estimation, and the data symbols are to transmit the information. The preamble part is made up of 4 symbols for AGC, 27-length Zadoff-Chu CAZAC (Constant Amplitude Zero Autocorrelation) sequence for synchronization, and 1 guard symbol. 27-length Zadoff-Chu sequence (k = 1) is defined as ck27   ck  0  ck 1  ck  26   and each element is as follows

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

 j 2 kn  n  1  ck  n   exp   , n  0,1, , 26 (1) 27   The preamble part is described in figure 2.

Figure 2. The preamble part configuration

2.2 Radio Modem Configuration In the implemented digital radio modem, QPSK modulation is used for achieving higher frequency efficiency than FSK. Moreover, we support AMC techniques to respond to a varying wireless channel condition. Our supporting AMC levels are shown in Table 1. Adding 16-QAM modulation to the AMC level and its implementation is considered as future work. Table 1. Used AMC Level

AMC 0 1

Modulation QPSK QPSK

Coding 0.75 1

The digital filters are configured for the narrow band digital radio at both transmitter and receiver sides. For the transmitter, digital filters are implemented by 64-interpolation RRC filter, 8interpolator, and 4-interpolator for converting 5 ksps symbol to 10.24 MHz system-clock signal. At the receiver, digital filter is configured with 64decimation matched filter and 2-decimation filter for 80 kHz sampling rate operation. We can get this sampling rate experimentally considering the tradeoff between hardware resource and performance from received signal processing gain.

2.3 Synchronization Implementation Scheme

We figure out that the 30 Hz is the maximum frequency offset for succeeding the synchronization with one correlator through measurement. Moreover, the frequency offset due to

133


thermal noise does not exceed ±120 Hz in our platform. Based on these results, we propose the following synchronizer.

within ±120 Hz are compensated and synchronized. The performance result will be shown at section 3.2. 3 MEASUREMENT OF THE SYSTEM PERFORMANCE

We conducted the performance measurement of the digital radio and synchronizer. On subsection 3.1, we describe environment for performing experiment and in sub-section 3.2, the results of measurement are shown for proving the superior performance of our implemented digital radio. 3.1 Measurement Environment

The measurement setup is shown in Figure 4.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

RF

LMR MODEM

RF

The proposed synchronizer is configured to nine correlators. The synchronization process has three steps in our proposed synchronizer. The first step is to generate nine pseudoreceived signals from the original received signal. The generated signals are located with 30 Hz spacing from each other. Such signals are obtained by multiplying the exponential term that indicates for frequency shift. The multiplying exponential terms for each correlator are same as Figure 3. In the second step, each correlator performs synchronization process using the respective input received signals, and the correlation values are calculated as a result. For the last step, the synchronizer finds the maximum correlation value from the results calculated in the second step and the correlator which has that maximum value. Finally, the received signal is selected as the one that is frequency-shifted according to the correlator with the maximum correlation value. Through these processes, the frequency offset of the received signals having a frequency

LMR MODEM

Figure 3. The synchronizer configuration

Figure 4. Digital radio performance measurement setup

Each digital radio platform is connected to PC. Chipscope pro of ISE14.7 is installed to analyze the performacne of synchronizer and to check constellation point. JTAG is used for the connection between PC and FPGA, and the agilent step attenuator is connected to the FER performance measurement according to the received power change. The 100 Hz frequency offset was intentionally applied for the performance test. In our platform, the conventional synchronizer with a single correlator does not work under 100 Hz frequency offset. However, our proposed synchronizer achieved three correlation metrics generation at 6th, 7th, and 8th correlators. The 7th correlator had the maximum value, and then 90 Hz frequency offset was compensated by

134


shifting all signals automatically. These results are described on next subsection. For the FER performance test of the digital radio, we measure the FER while the received sensitivity is changed from -107 dBm to -115 dBm using the step attenuator. Note that all the experiments were performed on 450 MHz UHF center frequency. 3.2 Measurement results

First of all, we measured the occupied bandwidth of the implemented narrow band digital radio. Our target bandwidth is within 5.2 kHz after passing through the digital filters. A measurement result using spectrum analyzer is presented in Figure 5, and this shows that our goal was achieved.

Figure 6. The performance of synchronizer under 100Hz frequency offset Table 2. The expectation receiver sensitivity calculation

Figure 5. The occupied bandwidth of digital radio

Next measurement step is the performance of proposed synchronizer under a 100 Hz frequency offset circumstance. The existing synchronizer with single correlator does not perform in this environment. However, the proposed synchronizer can generate the correlation metric value for 7th, 8th, and 9th correlators. Figure 6 shows the results of the synchronizer performance at -90 dBm, -100 dBm, and -110 dBm receiver sensitivity. These results were obtained using Chipscope pro and printed on PC monitor.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Item

Value

Unit

Thermal Noise Density

-174

dBm/Hz

Receiver Noise Figure

6

dB

Receiver Noise Density

-168

dBm/Hz

A

Receiver Noise Power

-130

dBm

A+10log10(6.25) (kHz)

Required SNR

12

dB

10-2 FER

Implementation Loss

6

dB

Receiver Sensitivity

-112

dBm

Finally, the FER of the digital radio is measured according to the receiver sensitivity change. The target received sensitivity is decided by both

135


calculation and simulation, and it is shown by the Table 2. The FER performance measurement results are as following. 0

10

AMC0: QPSK, 3/4 AMC1: QPSK, 1

5 ACKNOWLEDGEMENT

This work was supported by the ICT R&D program of MSIP/IITP, Republic of Korea. [13911-01-102, Developing the core technology of the digital LMR RoIP convergence wireless terminal with 6.25kHz channel bandwidth]

-1

10

FER

REFERENCES -2

10

[1]

FCC NAROWBANDING MANDATE A Public Safety Guide for Compliance, 2006.

[2]

ETSI TS 102 658 V2.3.1, Electromagnetic compatibility and Radio spectrum Matters; Digital Pribate Mobile Radio (dPMR) using FDMA with a channel spacing of 6.25 kHz

[3]

dPMR Whitepaper, “4 Level FSK/FDMA 6.25 kHz Technology”

[4]

NXDN Whitepaper, “4 Level FSK/FDMA 6.25 kHz Technology”

[5]

A. Shukla and Ravimohan, “Synchronization in Cognitive Radio Systems: A Survey,” Internation Jorunal of Advanced Research in Computer and Communication Engineering, vol. 3, Issue 7, pp. 7372-7375, July 2014.

-3

10

-4

10 -114 -113.5 -113 -112.5 -112 -111.5 -111 -110.5 Receiver Sensitivites(dBm)

-110 -109.5

-109

Figure 7. FER according to the received sensitivities

Our digital radio achieved the target FER of 10-2 at about -112 dBm. Besides, these results are attained without ARQ technique.

4 CONCLUSION

In this paper, we implemented the 6.25 kHz narrow band digital radio with 8,000 bps maximum data rate level. Especially, we proposed the synchronizer implementation scheme. Digital radio has been implemented on the FPGA-based platform which is developed by ourselves. A variety of measurements were performed through the testing instruments. The implemented digital radio overcomes performance degradation resulting from frequency offset and achieves 8,000 bps maximum data rate. The measurement results show that the synchronizer is working well within maximum ±120 Hz frequency offset, and the digital radio modem outperforms the conventional 6.25 kHz digital radio with 4,800 bps maximum data rates.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

136


LTE D2D - Challenges and Perspective of Mobile Operators Taisiya Kim¹, Soo Kyung Park¹ and Bong Gyou Lee² ¹Graduate Program in Technology & Business Administration, Yonsei University, Seoul, Korea. ²Graduate School of Information, Yonsei University, Seoul, Korea. {lucky8619, sk.park, bglee*}@yonsei.ac.kr

ABSTRACT Managing the tremendous data traffic is a main concern for mobile network operators in IoT environment. The growing of wireless network demand makes a burden to mobile network traffic and LTE D2D is regarding as a solution for the spectrum problem. LTE D2D will bring an impact on providers to share spectrum and on mobile environment. The main purpose of this study is to analyze the role and relationship of key players in changed mobile environment, caused by LTE D2D. The solutions for mobile operators were drawn through the literature review and in-depth interview of experts. The results of this study show that the profit source is going to become a main issue among key players and mobile operators. Thus, for mobile operators, it is important to measure communication details or data usage with certain software to prevent unilateral loss from free business model. KEYWORDS

mobile operators try to avoid the data traffic, by maintaining the wireless network or allocating the new frequency band [2]. Demand of wireless network keeps growing, but the limited resource of spectrum is assumed to cause overcrowded mobile network by the arrival of new services like Internet of Things (IoT). IoT is a new type of service that will share information through connection to mobile network of all things: home appliances, electronic equipment, automobiles etc. In order to invigorate IoT service, discovering new frequency band and developing new technology for prevention of data traffic congestion problem is important to solve the spectrum problem. While governments are taking initiative for discovering new frequency band, manufacturers try to develop new technology for increasing of spectrum capacity [3]. Mobile network operators predict Long Term Evolution Device to Device (LTE D2D) as a suitable solution to the spectrum problem for next generation. The discussions of LTE D2D were first held on 3rd Generation Partnership Project (3GPP) Onwards Workshop, and was

Long Term Evolution Device to Device (LTE D2D), Mobile Operators, Key Players, Spectrum Sharing

1 INTRODUCTION Smart devices such as smartphone, tablet PCs and other wireless connected devices diffuse rapidly. According to Ericsson, smartphone subscription rose up to 2.7 billion in 2014, and expected to keep increasing (Fig. 1) [1]. While number of subscribers is increasing, the information data is tremendously growing, and

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 1. Smartphone Subscriptions in 2014 [1]

137


Figure 2. Global Mobile Traffic in 2014 [1]

taken into consideration as a new candidate technology for 4G LTE Advanced Rel-12. In 2013, standardization works for requirements in Proximity-based Services (ProSe) were fixed and scenarios were defined[4]. The possible commercial scenarios show that LTE D2D can be an opportunity to those players, who plan to provide new services using LTE D2D. However it may fall to being used as a mere provider of dumb pipe for others, like mobile network operators. The issue such as network neutrality is possible to surface again [6]. The object of this study is to analyze roles of key players in LTE D2D commercial channel and to suggest opportunities to mobile network operators. Analysis on their role is referred to key players’ map of smartphone environment. First of all, the study describes issues related to key players, who have direct link with mobile operators in LTE D2D. Then, it accomplishes new relationships and needs of key players toward mobile operators by illustrating the conceptual map. And finally, proposes solutions to the issues among key players through expert interview of Korean mobile operators. 2 RESEARCH BACKGROUND Smartphone penetration rate in demographic terms of South Korea indicates 67.7%, which is the highest in the world [7]. It is 10% higher than Norway (55.0%), which is the 2nd place, and 4.6 times higher than global average. The reason of such high penetration and spread of service in South Korea attributes to the high

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 3. Network Platform Usage for Mobile Internet in South Korea [8]

quality network. According to Korea Internet & Security Agency (KISA), more than 86.9% mobile internet users in Korea access to LTE [8]. The high quality mobile network enables mobile operators to keep trying to develop new business models and services [9]. One of the main markets for next generation business is IoT. IoT connects things, share information, analyze it through mobile network, and communicate through intelligent system by itself. Mobile operators and ICT companies expect IoT to be means for next generation’s industry, but it is assumed to be saturated by tremendous demands of wireless communications. The main issue of Korean government is to procure sufficient frequency for broadband mobile services. Discussions on frequency shortage are under way not only in Korea, but also all over the world. Governments enforce new policies, such as distributing and sharing spectrum, and develop new technologies to deal with upcoming data traffic congestion. Korean government revised the Radio Regulation Law in 2014. The Law stipulates that mobile network operators are allowed to share frequency with others, such as ventures or small business operators. So far, frequency band in Korea has been exclusively used by broadcasters or mobile network operators. Now, as the Law allows frequency band to be used for ICT business or for social public services, it has created a proper environment for new service communications, such as LTE D2D. LTE D2D allows devices to communicate directly without using cellular network

138


infrastructure. It expects such an appropriate solution to decrease data congestion. Korea Telecom (KT Corporation) and Qualcomm announced an agreement to jointly develop LTE D2D [5]. KT Corporation is a leading operator that has already launched world’s first commercial LTE broadcast service. Thus, Qualcomm is planning to start pilot project in South Korea. LTE D2D seems to have more technical efficiency in the countries with high population density and high rated use of LTE device. Also, application of LTE D2D communication in the places with high population density can bring a significant added value. South Korea seems to have proper environmental condition for LTE D2D. Fig. 4 shows usage percentage of mobile data in main cities of South Korea. The capital, Seoul and Gyeonggi Province show the highest percentage of mobile data traffic congestion. Despite opportunities of LTE D2D, it can cause conflicts between mobile network operators, who have already been using frequency, and new players.

3 LTE D2D COMMUNICATIONS 3.1 Concept of LTE-D2D LTE D2D communication allows LTE based devices communicate without using cellular network infrastructure [4, 11, 12]. The devices communicate directly with one another when they are in close proximity. LTE D2D uses one resource, uplink or downlink, of frequency and LTE Time-Division Duplex frequency, which can decrease the massive data traffic. Currently, technologies like Bluetooth, NFC (Near Field Communications), and Wi-Fi Direct are using in proximity-based services. They are also sufficient for local-area communications, but it is hard to support mass of proximity-based services with abilities of these technologies. Bluetooth is made to transmit data to 100m range but in reality it is inferior. It is capable to communicate with low power consumption and mainly uses in peripheral devices, like headset and keyboard. The devices are proper for local area connection and not for high-volume multimedia communication services. Also it operates in unlicensed spectrum so it vulnerable to attack. Finally, because it uses different network from mobile, it has to parallel two networks and it increases the consumption of the battery [13, 14]. NFC has higher level of security than Bluetooth, however the coverage is only 10cm and also the communication speed is low. Therefore, it has restrictions on variety of business and uses in payment service.

Figure 4. Wireless Data Usage in Main Cities of South Korea [10] Figure 5. Leading Operators of LTE Broadcast [18]

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

139


Wi-Fi Direct also communicates directly between devices, like D2D does. However, D2D has wider coverage and possible to transfer more massive data than Wi-Fi Direct communication, because it uses LTE network. Wi-Fi Direct uses unlicensed wireless network and it is hard to control service or interruption from other device. It also discover device through two steps: device discovery and service discovery, but LTE D2D has only one discovery step. Process of LTE D2D: • Device discovery: discover the other LTED2D based device. • Link setup: setup link to connect two devices. • Data communication: transmit data through wireless link to the devices. LTE D2D communication process is simple compared to other communication technologies [14, 15]. The other important feature of LTE D2D is that it share the same spectrum with cellular device and spatial reuse it to improve system capacity. Through the next two steps, setting up link and transfer data between devices without network assisted information, LTE D2D will decrease amount of data traffic in the crowded places with many devices. Also LTE D2D saves power energy. When two devices are in close proximity they require lower transmission power level [17].

Figure 6. Comparison of LTE D2D with other D2D Communications [14]

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Beside technological advantages, LTE D2D is also a limelight in policy organizations. It became a candidate for IoT standardized technology and many application plans are discussing. 3GPP standardization plan is focused on two application channels of D2D communication service, discovery and communication channels. 3.2 LTE D2D Application Channels Communication channel enables mobile devices to communicate directly with each other and also secure connections by controlling the radio resource allocation. As US government expressed to use LTE for future public safety communications, many organizations expect that communication channel of LTE D2D is proper for public safety. Proximity-based services are attractive for public safety organizations, such as fire, ambulance, and police services [19]. 3GPP complied with a full range of requirements in supporting communication channel for national security and public safety [1]. Therefore it is not proper for business propose [14]. Discovery channel is going to be used for commercial business. Discovery channel permits a mobile device to look for the other device and ascertain information about it [19].

Figure 7. Process of LTE D2D [15, 16]

140


players of the study are mobile operators, hardware & chip vendors, and service & contents providers. The issues of the key players are descripted bellow.  Mobile Operators

Figure 8. Services Capable in LTE D2D [20]

For example, contents provider can send necessary information to customer nearby, send commercial interests, or match information with advertisement. LTE D2D provides a common language for discovery; it operates horizontally across apps, operating systems, devices, and operators, exponentially expanding the field of value for proximal discovery. It is also expected to be used in many other services, such as information delivering service GeoFencing, network games, social matching, and targeting advertisements in the future [20]. 4. CHANGE OF KEY PLAYERS’ ROLE AND ANALYZING THEIR NEEDS The flow of mobile service ecosystem is illustrated in Fig. 9. Avrind (2009) classified stakeholders into four types identified by questions ‘who’, ‘what’, ‘how’, ‘how else’ and illustrated the value flow types [21]. Based on this classification, this study analyzes roles of key players of LTE D2D communication ecosystem, focused on mobile operators’ perspective.

LTE network is appropriate for transmitting multimedia contents that could cause mass of data traffic. In mobile operators’ perspective, LTE D2D will decrease the maintenance costs of wireless network. In addition, by expanding coverage, data provision in coastal areas and mountainous regions becomes possible in stable speed. As number of mobile phone is higher than population in South Korea, mobile operators compete with transmission and reception speed of the data. Hence, with LTE D2D, they can utilize the acceleration of data speed for marketing and can have competitive advantage in the LTE quality competition and development competition for next 5th Generation (5G) mobile technology. Mobile operators, however, may face two major problems; first, data usage of the users may decrease in the short term and Average Revenue per User (APRU) may be reduced in the long term [9]. Second, mobile operators may lose their control and domination of the network, because information is not stored in the communications company as the communications between devices do not pass through base stations [14].

4.1. Identification Issues of Key Players The key players and their issues in LTE D2D commercial channel are identified by reviewing papers, articles, news and studies, related to the environment [4, 11, 12, 13, 14, 18, 19, 22, 23]. Through the literature review, the study sets key players, who directly are going to be related with mobile operators. The final key

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 9. Smartphone Ecosystem of Key Payers [21]

141


 Hardware & Chip Vendors The hardware & chip vendors play an important role in LTE D2D. They can develop devices with embedded technology or just insert a chip into existing devices. Depends on the players, the diffusion method of the technology will be change. Therefore mobile operators and hardware & chip vendors should cooperate. KT Corporation and Qualcomm started LTE D2D development project together from 2014 and calls it ‘LTE Direct’. They are planning to start pilot project in South Korea. When LTE D2D is commercialized, hardware & chip vendors will have the benefit of expanded market. Thus, the development of chips must fit their scale of profit and loss. If the government allows LTE D2D only to be used for public disasters, hardware & chip vendors will have difficulty in making profit with disaster network because it is estimated that the demand for disaster mobile phone remains 200,000 units in Korea [5]. Therefore, hardware & chip vendors will expect to conduct new commercial service such as advertisement by using discovery channel of LTE D2D. Especially, companies like Qualcomm can use the framework among devices in the future business, such as for IoT environment [23].  Service & Contents Providers In this study, service & contents providers are including not only content owners/creator, mobile app developers but also advertisers [21]. If service and contents providers use discovery channel, all device users in 1km could become targets for promotion, and they will become advertisers. Location-based Service (LBS) providers and mobile app developers can participate in this kind of business. Various businesses are engaged in marketing activities such as coupon mailing by using location information of the smart phone users. In the past, this kind of target marketing required passing through specific service platforms. In the case of LTE

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

D2D, information transmission is possible through built-in chips without passing through service platforms. Thus, small businesses or advertisers can deliver their information in cheap cost. Various businesses can participate in the business utilizing LTE D2D by using the network of the mobile operators. As service & contents providers engage in business by utilizing the network of mobile operators, there may be cases when their business models overlay each other. 4.2. Role and Relationships of Key Players To compare pre and post introduction of mobile environment, the study developed Avrind’s mobile ecosystem framework. Subjects of the relationship analysis were limited to mobile operators, hardware & chip vendors, and service & contents providers. For the parsimony of the study, other players were excluded. In the past, there was clear distinction between mobility provider and mobility contents. As section 4.1 reviewed, now mobile operators, hardware & chip vendors, service & contents providers can all provide mobility contents. In this case, conflicts may arise since business model can overlap one another. Borrowing the study of Cameron, B. G., et al. (2008), this study classifies the needs of players into 1) common needs, 2) synergistic needs, 3) conflicting needs and 4) orthogonal needs [24].

Figure 10. Conceptual Map of LTE D2D Key Players

142


The roles of key players, which change around mobile operators in LTE D2D environment, are shown in Fig. 10. 4.3. Directions of Korean Mobile Operation Experts This study verified the conceptual map presented in Fig. 10 through literature review and conducted expert interviews to collect opinions. Respondents of the interview were 9 experts, who work for major mobile operators of South Korea, SKT, KT, LG U+. They were restricted to experts in strategy department or research department with minimum 12-year career experience in the field. Represents that has prior knowledge on LTE D2D were selected. Questions were composed of the strengths and weaknesses of LTE D2D in the perspective of mobile operators, verification on the conceptual map and prediction of relationship with key players. The results can be summarized as follows; mobile operators can maintain the stable wireless network with reduction of traffic burden. Also, new business models are available, such as IoT. There was prediction that if LTE D2D service is provided by mobile operators, it may be beneficial to the maintenance of subscribers with the expansion of conveniences. Still, most discussions about strengths and weaknesses focused on the traffic and profitability. For the investment of mobile operators in frequency and access network, realistic source of profit is the price users pay for their usage of traffic (Korea mobile operators paid a total of US 2.4 billion dollars in the auction for LTE frequency). Therefore, although the commercialization of LTE D2D is desirable in term of cost reduction due to decreased traffic burden, it might not be beneficial to profit-making, if it becomes difficult to bill on the traffic. Especially, as current billing system or communication system cannot accommodate circulation of information which does not pass through base

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

station, this can work as a burden for development of new technology and investment. Then, conceptual map was conducted by considering the relationship between the mobile operators and other key players. At the moment, the relationship between mobile operators and hardware & chip vendors has synergistic needs. There were opinions, however, that they can develop into a relationship with conflicting needs depending on business model. In addition, in Korea, mobile operators have bargaining power in buying the devices. Hence, there is little possibility for hardware & chip vendors to load functions unfavorable for mobile operators. Therefore, if hardware & chip vendors make profit-making structure in cooperation with mobile operators, they can spread devices much quickly. The relationship between mobile operators and contents provider are as follows; first, if the contents provider can create paid business model and second, if LTE D2D-based platform is attractive as an advertising medium. In order for LTE D2D service to become an advertisingbased business model, securing sufficient number of subscribers is the key for successful entry which can be recognized as stable advertising base by the advertisers. Hence, both securing early subscribers (dissemination rate of device) and creation of ecosystem are important. The most important issue between mobile operators and contents provider was billing system as well. As service & contents providers use mobile operators' network, there has be the issue of network neutrality. Network neutrality started from the structure in which users' traffic continuously increases in unlimited environment while the network operators bear all the burden of investment in the network. Although some experts had the opinion that commercialization of LTE D2D would cause similar problems, it was predicted that the discussion might be different from existing network neutrality because mobile operators cannot argue their rights (including the burden of traffic) other than use of frequency as LTE

143


D2D conducts direct communication without passing through base stations of mobile operators. In addition, there were worries that in case users of unlimited data cause traffic with the role of gateway or offload through landline, traffic and profit structure of the operators can be changed, which might be a difficult condition for the mobile operators. In addition, the questions were raised as to whether additional investment is needed for the security of contents which do not pass through network and who’s going to make the investment. Solutions for the issues discussed so far were suggested as follows; for LTE D2D to be profit source for the mobile operators, it is necessary to find ways to measure communication details and amount of data usage by using software. If mobile operators lose control after the commercialization of LTE D2D, they only provide their network and don’t get any benefit from it. Although there is possibility for the issue of network neutrality to be raised based on the technology method, it was predicted that the problem is economically soluble. It is an important task for mobile operators to prevent unilateral loss from free business model. 5. CONCLUSION This study discussed issues which can take place when LTE D2D commercial channel is utilized and their solutions in terms of mobile operators. Through literature review, the study selected key players of hardware & chip vendors and service & contents providers and presented analysis of their needs and relationship with a map. Study presented Avrind’s (2009) ‘Smartphone Ecosystem of Key Payers in Conceptual Map’ in illustrating the LTE D2D environment. Conceptual map shows the value of key flow with a map and analyzes their needs. Interviews were conducted on the experts working in mobile operators to verify the conceptual map and collect their opinions. As the result of analysis, it is desirable for hardware & chip vendors to

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

cooperate with mobile operators as mobile operators have influence on the distribution of mobile devices. This relationship should spread devices and create profit structure in synergistic cooperative relationship. In order for LTE D2D to be an attractive service, the roles of mobile operators and contents provider are very important, which will decide whether or not LTE D2D will be a successful advertising platform. The most important issue, however, is none other than billing system. For LTE D2D to be a profit source for mobile operators, it was analyzed that an alternative is required which can measure the communication details and data usage. As discussed, for successful dissemination of LTE D2D, it is important for the key players to effectively build this service ecosystem. Up until now, LTE D2D technology has been used on simple areas such as connection to accessories but it is expected that it will be used in various fields such as network games, social matching service and target advertisement. Therefore, whether it will be an essential technology or will disappear depends on the roles of these key players. This study has its limitations as it was conducted at the early discussion stage of LTE D2D when there was no business model. Still, this study has its meaning in conducting the perspective of mobile operators, who are the key players. It provides directions and discusses solutions for the issues among players, to help LTE D2D service take firm place as an essential service in the future. ACKNOWLEDGEMENT This research was supported by the MSIP (Ministry of Science, ICT and Future Planning), Korea, under the ITRC (Information Technology Research Center) support program (NIPA-2014-H0301-14-1042) supervised by the NIPA (National IT Industry Promotion Agency)

144


REFERENCES [1] R. Qureshi, “Ericsson Mobility Report”, Ericsson, pp. 1-32, 2014. [2] Y.J. Lee, J. Oh, and B.G. Lee, “Logical Push Framework for Real-time SNS Processing”, Proceedings of The 4th IEEE International Conference on Computational Aspects of Social Networks, pp. 47-51, 2012. [3] E.S. Jun, K.C. Park, and B.G. Lee, “Analyzing Spectrum Management Policy for Utilizing TV White Space”, Proceedings of The International Conference on System Electronic Engineering, 2012. [4]

Korea Communiactions Agency(KCA), “D2D Technology and Market Trends”, Korea Communiactions Agency , Issues & Propection, vol. 16, pp. 1-10, 2013.

[5] Edaily News, “Is Qualcomm going to start deviceto-device communication era?”, 2015. Available at: http://www.edaily.co.kr/news/NewsRead.edy?SCD= JE31&newsid=01115206609234768&DCD=A0050 3&OutLnkChk=Y [6] J.Y. Kim, J.K. Park and B.G. Lee, “Mobile Network Neutrality in Smart Phone Era”, Proceedings of The ITS Asia-Pacific Regional Conference, 2010. [7] Money Today News, “Generation of 40 million smartphone subscribers is coming”, 2014. Available at:http://www.mt.co.kr/view/mtview.php?type=1&n o=2013122423081279407&outlink=1 [8] Korea Internet & Security Agency(KISA) “Survey on mobile internet use in 2014”, pp.1-23, 2014. [9] S.K. Park, J.H. Kwak, and B.G. Lee, “Is ARPU the Right Choice for Wireless Data-based Communication Services?”, Proceedings of The 4th IEEE International Conference on Computational Aspects of Social Networks, 2012. [10] Korea JoongAng Daily, “If you know data traffic, you can get future purchising power”, 2011. Available at: http://article.joins.com/news/article/arti cle.asp?totalid=6793758&cloc=olink|article|default

[14] T.J. Kim, “D2D, Changes and Implications”, Digieco, Issue&Trend, 2014. [15] S.I. Sung, J.W. Hong, J.S. Kim, S.I. Park, C.W. Park, S.H. Choi and K.B. Lee, “Cellular Networks D2D Communications Trends”, Journal of The Korean Institute of Communication Sciences, vol. 29, no. 7, pp. 97-105, 2012. [16] J.W. Hong, S.H. Choi, S.I. Sung, K.B. Lee, S.I. Park, C.W. Park and J.Y. Kim, “D2D Communications Technology and Standardiztion”, The Magazine of the IEEE, pp. 371-381, 2013. [17] G.M. Kang, H.M. Kim, J.H. Nam and O.S. Shin, “LTE-Advanced Communications and Techonology Standardiztion”, Journal of Electromagnetic Engineering and Science, vol. 24, no.5, pp.52-64, 2013. [18] Qualcomm, “LTE Direct: Operators Enabled Proximity Services”, 2014. [19] A. Brydon, “Opportunities and Threats from LTE Device-to-Device Communication”, Available at: http://www.unwiredinsight.com/2014/lte-d2d [20] Multilog Blog, “Next device-to-device direct communication, LTE Direct, Concept and view”, 2014. Available At: http://donghun.kr/1375 [21] A.S. Arvind, Stakeholder Value Network Analysis for the Mobile Services Ecosystem, Doctoral Dissertation, Massachusetts Institute of Technology, 2009. [22] L. Lei, Z. Zhong, C. Lin, and X. Shen, “Operator Controlled Device-to-device Communications in LTE-advanced Networks”, IEEE Wireless Communications, vol. 19, no. 3, pp. 96-104. 2012. [23] Qualcomm, “LTE Direct Always-on Device-toDevice Proximal Discovery”, 2014. [24] B.G. Cameron, E.F. Crawley, G. Loureiro and E.S. Rebentisch, “Value Flow Mapping: Using Networks to Inform Stakeholder Analysis”, Acta Astronautica, vol. 62, no. 4, pp. 324-333, 2008.

[11] K. Doppler, M. Rinne, C. Wijting, C. Ribeiro, and K. Hugl, “Device-to-device Communication as an Underlay to LTE-advanced Networks”, Communications Magazine, IEEE, vol. 47, no. 12, pp. 42-49, 2009. [12] A. Asadi, and V. Mancuso, “WiFi Direct and LTE D2D in Action”, IEEE, In Wireless Days, pp. 1-8, 2013. [13] S. Mumtaz, and J. Rodriguez, “Introduction to D2D Communication”, Smart Device to Smart Device Communication, Springer International Publishing, pp. 1-22, 2014.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

145


Performance Analysis of Cross Component Carrier Scheduling in LTE Small Cell Access Point System Sangchul Oh, JeeHyeon Na and Dongseung Kwon Wireless Application Research Department Electronics and Telecommunications Research Institute (ETRI) 218 Gajeongno, Yuseong-gu, Daejeon, 305-700, KOREA E-mail: [email protected], [email protected], [email protected] ABSTRACT With respect to the carrier aggregation (CA) in a LTE small cell access point system, the performance of an independent and cross component carrier (CC) scheduling is simulated and analyzed in this paper. According to the results, we obtained a capacity increase of about 17~46% for the cross CC max rate (MR) algorithm and about 12~45% for the cross CC proportional fair (PF) algorithm. On the other hand, in the round robin (RR) scenario, we cannot notice the differences between the independent CC and cross CC scheduling policies on allocating resources. We also found that CA frequency combination 1.8GHz CC + 800MHz CC has a better performance when comparing to the other combinations with respect to total average UE throughput. KEYWORDS LTE, Small Cell Access Point, Carrier Aggregation, Cross Carrier Scheduling.

1 INTRODUCTION A small cell access point (AP) is a nomadic or mobile access point with small size supporting a cheap wired and wireless convergence service by connecting a mobile phone with an internet in an indoor environment such as home and office. Although it is similar to a Wi-Fi access point in a functional aspects point of view, it is different that a primary role of a small cell

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

access point is to relay a mobile phone call unlike a Wi-Fi AP. It is an evolved technology in a view point of serving an internet as well as a voice unlike a legacy wired and wireless convergence service such as Onephone and Homezone services. In a long term, a small cell access point based on long term evolution (LTE) will become a practical application of integration of a voice and data service in a home and office. It is also able to recognize an exact location and an indication of coming or going of their families through an individual small cell access in terms of a location-based service (LBS), and to provide new supplementary services by combining various home local area network (LAN) technologies. 3GPP has attempted to find a new technology that can provide both higher data-rate support in a wider bandwidth (BW) with an LTE system extended up to 100MHz, and backward compatible co-existence in LTE systems with an aggregation of multiple component carriers into an overall wider bandwidth. Carrier aggregation (CA) is an important technology that combines multiple radio channels within and across bands to increase the user data rates and reduce latency. Two or more component carriers (CCs) are aggregated in an LTE system to support wider bandwidths of up to 100 MHz. This feature allows the scalable expansion of the effective bandwidth delivered to a user terminal through the concurrent utilization of radio resources across multiple carriers. These carriers may be of

146


different bandwidths, and may be in the same or different bands to provide maximum flexibility in utilizing the scarce radio spectrum available to operators. Support for this feature requires an enhancement to the PHY, MAC, and RRC layers of LTE Release 8 and Release 9 while ensuring that LTE Release 10 preserves backward compatibility with the LTE Release 8 and Release 9 specifications. As a matter of fact, LTE user equipment (UE) can simultaneously receive or transmit on one or multiple CCs aggregated up to 100MHz, but Release 8 and 9 UE can only receive and transmit on a single CC. According to the 3GPP Release 10 specifications in [1]-[4], all CCs shall be LTE Release 8 and 9 compatible, but existing mechanisms such as barring may be used to avoid Release 8 and 9 UEs to camp on a CC. Moreover, CA have been required in small cell AP to maximize system capacity with both macro base station and small cell AP as well. According to [5], in practice, spectrum allocation for an operator is often dispersed along the frequency bands with large frequency separation. Hence heterogeneous CC is considered in this paper. In uplink resource allocation aspects, [6] claimed the 2 CC simultaneous access scheme outperforms CC selection and performs almost equally to the mCC simultaneous access scheme (m≥3). On the other hand, with respect to the downlink CA in a LTE system, a downlink cross component carrier (CC) scheduling algorithm is analyzed in this paper. The performance of the cross CC scheduling algorithm on heterogeneous CCs with different channel characteristics has been compared through computer simulations. The rest of this paper is organized as follows. Section 2 describes the small cell AP architecture considered in this paper. The LTE protocol stack interaction for CA is mentioned in section 3. Details of the scheduling algorithm for carrier aggregation are presented in section 4. We then provide the simulation

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

environments and results in section 5, and offer some concluding remarks in section 6. 2 SMALL CELL AP ARCHITECTURE Figure 1 shows a logical interface for the small cell AP that has a set of S1 interfaces to connect the small cell AP to the EPC [1]. The evolved universal terrestrial radio access network (EUTRAN) consists of the macro eNode B (eNB), the home eNode B (a.k.a. small cell AP) and the evolved packet core (EPC) corresponding to the mobility management entity (MME) and the serving gateway (SGW). The E-UTRAN architecture may deploy a small cell AP gate way (GW) to allow the S1 interface between the small cell AP and the EPC to scale to support a large number of small cell APs. The small cell AP GW serves as a concentrator for the control-plane (C-Plane), specifically the S1-MME interface. The S1-U interface from the small cell AP may be terminated at the small cell AP GW, or a direct logical user-plane (U-Plane) connection between small cell AP and SGW may be used. The small cell AP GW appears to the MME as an eNB. The small cell AP GW appears to the small cell AP as an MME. The S1 interface between the small cell AP and the EPC is the same whether the small cell AP is connected to the EPC via a small cell AP GW or not. The small cell AP GW shall connect to the EPC in a way that handover to cells served by the small cell AP GW shall not necessarily require interMME handovers. One small cell AP serves only one cell. The functions supported by the small cell AP shall be basically the same as those supported by a macro eNB and the procedures run between a small cell AP and the EPC shall be the same as those between an eNB and the EPC.

147


MME / S-GW S1

S1

S1

S1

S1

S1

S5

S1

S1

HeNB GW

X2 S1

X2

S1

X2

eNB

Figure 3 presents LTE protocol interactions to

E-UTRAN

eNB

eNB

3 LTE PROTOCOL STACK INTERACTION FOR CARRIER AGGREGATION

MME / S-GW S1

MME / S-GW

X2 X2

X2

X2

HeNB HeNB

HeNB

Figure 1. Small Cell AP Interface.

2 shows some of the potential deployment scenarios for CA. Scenario #1 is that F1 and F2 cells are colocated and overlaid, providing nearly the same coverage. Scenario #2 is that F1 and F2 cells are co-located and overlaid, but F2 has smaller coverage due to larger path loss. Scenario #3 is that F1 and F2 cells are co-located but F2 antennas are directed to the cell boundaries of F1 so that cell edge throughput is increased. Scenario #4 is that F1 provides macro coverage and on F2 remote radio heads (RRHs) are used to improve throughput at hot spots. Scenario #5 is similar to scenario #2, but frequency selective repeaters are deployed so that coverage is extended for one of the carrier frequencies.

Figure

(2) scenarios #2

(1) scenarios #1

(3) scenarios #3

(4) scenarios #4

Macro eNB

(5) scenarios #5

Small Cell AP

F1

Figure 2. Carrier Aggregation Deployment Scenarios (F2 > F1).

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

F2

support the CA. As mentioned in section 1, support for CA requires an enhancement to the PHY, MAC, and RRC layers of LTE Release 8 and Release 9. A device capable of CA has one downlink (DL) primary component carrier (PCC) and one associated uplink (UL) PCC. Basic linkage between the DL and UL is signaled in SIB Type 2 described in [3]. The device may have one or several secondary component carriers (SCCs) added in RRC CONNECTED mode only. The configuration of PCC is not cellspecific but UE-specific. The handover is performed using a PCC, and the network may decide to switch the PCC for another device using a handover procedure. Radio resource management (RRM) application controls all procedures and configurations of these protocol layers such as PHY, MAC, and RRC regarding CA. The basic role of RRM is to ensure that the radio resources are efficiently used, taking advantage of the available adaptation techniques, and to serve the users according to their configured Quality of Service (QoS) parameters. 3GPP specifies the RRM related signaling but the actual RRM algorithms in the network are not defined in 3GPP. Those algorithms can be dependent on vendor and operator. A UE can be configured by radio resource control (RRC) message on non-configured CCs as well as configured CCs. The configurations of the PCell and SCell are performed by RRM control. A UE has only one RRC connection with the network. PCell indicates a cell configured on a PCC, and SCell indicates a cell configured on an SCC. When adding a new SCC, dedicated RRC signaling is used to send all required system information of the SCC. PCC can only be changed using a handover procedure (i.e., with

148


CC1 (PCC)

PDCCH

CC2 (SCC)

PDSCH

PDSCH

PDSCH

CC2 (SCC)

PDCCH

CC1 (PCC)

1 msec subframe

PDCCH

1 msec subframe

PDCCH

a security key change and RACH procedure). PCell provides non-access stratum (NAS) mobility information for an RRC connection establishment, re-establishment, or handover. A re-establishment procedure is triggered when PCell experiences a radio link failure (RLF), but not when SCells experience such a failure. Data received from an RLC is aggregated in MAC and distributed into individual HARQ processes.

PDSCH

(1) Independent carrier scheduling

(2) Cross carrier scheduling

Figure 4. Resource Scheduling with Downlink Control Channel Signaling

4 SYSTEM MODEL

Figure 3. LTE Protocol Stack Interaction for Carrier Aggregation

Resource scheduling with downlink control signaling is shown in Figure 4. Resource can be assigned to a UE in two ways as follows. Independent carrier component scheduling separates PDCCH for each CC and reuses PDCCH structure and DCI formats in Release 8 and 9. On the other hand, cross component carrier scheduling shares common PDCCH for multiple CCs and reuses PDCCH structure in Release 8 and 9. It have new 3-bit carrier Indicator Field (CIF) added to DCI in Release 8. PCC cannot be cross scheduled, it is always scheduled through its own PDCCH. Main motivation of cross component scheduling was load balancing and interference management for control channels in heterogeneous networks.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

The perfect channel state information (CSI) is assumed at both the UE and small cell AP. Each subcarrier within all CCs can be used by only one user at any given time. CC and subcarrier allocation are performed at the small cell AP, and the users are notified of the CCs and subcarriers chosen for them. A system with K users, N component carriers (CCs), M subcarriers, and the time divided into time slots is considered. At each time slot, each scheduled user k will transmit on the allocated CC and subcarrier. Equal power and equal bandwidth allocation algorithms in all CCs are taken into consideration, which simply distribute the transmission power and bandwidth equally among the CCs and subcarriers. The total average UE throughput, E[ Rkn,m (t )] , can be formulated as follows.





E Rkn,m t  = 2  n  P H k ,m t  log 2 1    N0B N 2 KMT n 1k 1m1 t 1 M n  

B

N K M T

1

    

(1)

where E[ Rkn,m (t )] indicates the total average UE throughput on every user, which means the small cell AP average system capacity. Rkn,m (t )

149


indicates the instantaneous data rate of each user at each CC, and for each subcarrier and time slot. N is the total number of CCs, K is the total number of users, and M is the total number of subcarriers. T indicates the total number of time slots, and M n indicates the total number of subcarriers belonging to the nth CC. B and P indicate the total bandwidth and power. H kn,m t  indicates the channel gain on the nth CC, mth subcarrier of the kth user, and tth time slot. N 0 is the power spectral density of the AWGN. Round robin (RR), max rate (MR), and proportional fair (PF) scheduling that are representative algorithms among traditional scheduling policies is considered in this paper [7]-[8]. Two different types of scheduling, the independent CC scheduling and cross CC scheduling for carrier aggregation (CA), were compared for system resource allocation. In addition, RR, MR, and PF scheduling applying the independent CC scheduling algorithm and cross CC scheduling algorithm policies were evaluated to compare the simulation results from the average UE throughput. 5 SIMULATIONS Component carrier characteristics and CA frequency combination applied to simulation are described in Table 5 and Table 6 respectively, the results of which are shown in this section. The simulations assume that equal power and equal bandwidth allocation algorithms in all CCs are taken into consideration, and the transmission power and bandwidth are distributed equally among the CCs and subcarriers. This paper simulates an urban environment where the users are walking at 3 Table 7 describes the simulation km/h. environments using the detailed system parameters and assumptions in this paper. The equations (1) is used for the simulation and performance evaluations between the

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

independent CC and cross CC scheduling algorithm. Table 5. Component Carrier Characteristics.

Carrier frequency

Bandwidth

Carrier type

2.1GHz 1.8GHz 900MHz 800MHz

20MHz 20MHz 10MHz 10MHz

PCC PCC SCC SCC

Table 6. CA Frequency Combination.

CA Combination Index

CA frequency combination

A B C D

2.1GHz CC + 800MHz CC 2.1GHz CC + 900MHz CC 1.8GHz CC + 800MHz CC 1.8GHz CC + 900MHz CC

Table 7. System Parameters.

Parameters

values

AP Tx Power Traffic Type MAC Packet scheduling Maximum number of users tc Total time slot mobile speed

20 dBm Full buffer RR, MR, PF 32 1000 [9] 2048 3 km/h

As can be seen from the figures, CA combination index C in Figure 10 has a better performance compared to the other combinations with respect to total average UE throughput. This information indicates that CA frequency combination with a lower frequency band exhibits better channel gain than combinations with a higher frequency band. Hence, we found that CA frequency combination with a lower frequency band should be taken into account with high priority when performing CA. Figure 10 shows the simulation results of CA combination index C. As we can see in the figure, the cross CC MR and PF algorithms

150


6

5.5

x 10

6

5.5

x 10

Total Average UE Throughput (PCC:SCC=2.1GHz:0.9GHz)

Total Average UE Throughput (bps)

5 4.5 4 3.5 MR-indep-CC PF-indep-CC RR-indep-CC MR-cross-CC PF-cross-CC RR-cross-CC

3 2.5 2 1.5 1

0

5

10

15

20

25

30

35

Users

Figure 9. Total Average UE Throughput of CA Combination Index B. 6

5.5

x 10


5 Total Average UE Throughput (bps)

outperform the independent CC MR and PF algorithms. We obtained a capacity increase of about 17~46% for the cross CC MR algorithm and about 12~45% for the cross CC PF algorithm. However, the scheduling gain (i.e. gap) between the cross CC PF and independent CC PF algorithms is decreased consistently when the number of simultaneous users is increased, since the PF algorithm was designed basically for maximizing the system throughput and maintaining fairness. On the other hand, in the RR scenario, there are no apparent differences between the independent CC and cross CC scheduling policies in terms of allocating resources, since all users have the same average channel response within all CCs in the RR algorithm, which allocates all subcarriers of each CC to one user at each time slot independently of the user’s channel response. Plots on Figure 8, Figure 9, and Figure 11 have same pattern with Figure 10 in total average UE throughput aspect. It means that scheduling policy is not in association with CA frequency combination. Total Average UE Throughput (PCC:SCC=2.1GHz:0.8GHz)

4.5 4 3.5

MR-indep-CC PF-indep-CC RR-indep-CC MR-cross-CC PF-cross-CC RR-cross-CC

3 2.5 2 1.5 1

5

0

5

10

15

20

25

30

35

4.5

Figure 10. Total Average UE Throughput of CA Combination Index C.

4 3.5


3 2.5 2

6

5.5

1.5 1

0

5

10

15

20

25

30

Users

Figure 8. Total Average UE Throughput of CA Combination Index A.

x 10


5

35



Users

4.5 4 3.5


3 2.5 2 1.5 1

0

5

10

15

20

25

30

35

Users

Figure 11. Total Average UE Throughput of CA Combination Index D.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

151


6 CONCLUSION In this paper, we discussed important properties that take place in a carrier aggregation environment of an LTE system in which the system resources are shared among users over all CCs, and where their channel responses change independently on heterogeneous CCs. To analyze and evaluate the performance of cross component carrier scheduling considering different scheduling policies such as MR, PF, and RR algorithms, wireless urban environments where users are walking at 3km/h were simulated. The performance of the independent and cross CC scheduling algorithms has been analyzed and compared through computer simulations as well. As we can see in the simulation results of section 5, CA combination index C has a better performance when comparing to the other combinations with respect to total average UE throughput. This information indicates that CA frequency combination with a lower frequency band exhibits better channel gain than combinations with a higher frequency band. Hence, we found that CA frequency combination with a lower frequency band should be taken into account with high priority when performing CA. Furthermore, we observed a capacity increase of about 17~46% for the cross CC MR algorithm and about 12~45% for the cross CC PF algorithm. On the other hand, in the RR scenario, we cannot notice the differences between the independent CC and cross CC scheduling policies on allocating resources, since all users have the same average channel response within all CCs in the RR algorithm, which allocates all subcarriers of each CC to one user at each time slot independently of the users’ channel response. We leave the study of the cross CC scheduling performance of various modulations and coding techniques over wireless channels for further research.

This research was funded by the MSIP (Ministry of Science, ICT & Future Planning), Korea in the ICT R&D Program 2014. REFERENCES [1]

[2]

[3]

[4] [5]

[6]

[7]

[8]

[9]

3GPP, “E-UTRA and E-UTRAN Overall description; Stage 2 (Release 10),” TS 36.300, V10.4.0, June, 2011. 3GPP, “E-UTRA Medium Access Control (MAC) protocol specification (Release 10),” TS 36.321, V10.2.0, June, 2011. 3GPP, “E-UTRA Radio Resource Control (RRC); Protocol specification (Release 10),” TS 36.331, V10.2.0, June, 2011. 3GPP, “E-UTRA Physical layer procedures (Release 10),” TS 36.213, V10.2.0, June, 2011. W. Hua, R. Rosa, and K. Pedersen, “Performance Analysis of Downlink Inter-Band Carrier Aggregation in LTE-Advanced,” Proc. VTC Fall IEEE, Sept. 2011. pp. 1-5 Y. Choi, Y. Lee, K. Chang, “Adaptive Resource Allocation for Uplink Carrier Aggregation Scheme in LTE-A-Type Networks,” ETRI Journal, Volume 34, Number 5, October 2012, pp.759-762 E. Dahlman, S. Parkvall, J. Sköld, et al., “3G Evolution HSPA and LTE for Mobile Broadband Second edition,” Elsevier, Ltd. 2008. pp. 105-123. T. D. Nguyen and Y. Han, “A Proportional Fairness Algorithm with QoS Provision in Downlink OFDMA Systems,” IEEE Communications Letters, vol., 10, no. 11, Nov. 2006, pp. 760-762. A. Jalai, R. Padovani, and R. Pankaj, “Data Throughput of CDMA-HDR a High Efficiency-High Data Rate Personal Communications Wireless System,” in Proc. IEEE Vehicular Technology Conf. 2000-Spring, vol., 3, May 2000, pp 1854-1858

7 ACKNOWLEDGMENT

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

152


Novel Wireless Lighting-Control System with a Wireless DMX Link for LargeScale Light Shows Eun Ho Kim Korea Institute of Industrial Technology 143, Hanggaul-ro, Sangrok-gu, Ansan-si, Gyeonggi-do, SEOUL, 426-910 Rep. of KOREA [email protected] ABSTRACT Systems for large-scale light shows require a large amount of electrical wire, leading to problems with the mobility of the control unit and the maintenance of the lighting system. A wireless DMX link between the main controller and the light controllers (e.g., dimmers) would be instrumental in overcoming these constraints. Previously developed wireless DMX links are typically limited to smallscale shows with small audiences. Because the audience uses wireless devices, data transmission by wireless DMX link cannot be guaranteed in realtime. Hence, in this work, a novel wireless lightingcontrol system with a wireless DMX link for largescale light shows is proposed. The proposed wireless DMX link can transmit data in real time, and store data beforehand, which can called upon during the light show with simple wireless commands.

KEYWORDS Wireless DMX, DMX512 protocol, LED, Light Show, IEEE 802.

1 INTRODUCTION Recently, LED lights have been used in various fields and especially in the arts, lights festivals, theme parks, and outdoor facilities, in which the scenery is colored with LED lights. The availability of these devices, along with their low-energy consumption, long life span, and ease of use, have helped LED lights to emerge as the next generation of lighting devices, replacing existing halogen and fluorescent lighting [1]. Therefore, the demand for LED lights in large-scale light festivals and

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

performances is on the rise, and research is actively being conducted into such applications. A lighting system controls multiple lights by establishing communication through tools such as Power Line Communication (PLC), Digital Addressable Lighting Interface (DALI), Digital MultipleX 512 (DMX512), and CAN [2]. Among these tools, the DMX512 communication protocol is a popular choice to manage effects and lighting in theaters, on stage, and at concerts. Various stage effects can be created with the device beyond merely controlling the lights, such as manipulation, programing, pan/tilt, shutter, and timer, and this breadth explains why the device has become widely used in these fields. It is simple to move, and easy to install and use. Moreover, the control features are reliable, and the wireless DMX512 system has made substantial developments towards the technology required in large-scale performances, gradually increasing the system’s effectiveness. Although there are benefits to wire-based lighting-control methods, such as accurate data delivery and the ability to supply power, the structure of the stage, along with the installation and maintenance of additional equipment, is constrained in such a system. By comparison, lighting-control methods that use wireless technology have considerable advantages over wire-based methods insofar as they overcome the biggest drawbacks of such systems: installation and maintenance. However, the disadvantage to a wireless-based method is that wireless communication is inherently unstable compared with wired technology. Nonetheless, in practice, lights are being directed on a large

153


scale with wireless technology, despite these stability issues, and this signifies a need to develop a wireless-based DMX512 system that can provide stability. Existing wireless-based DMX (WDMX) controllers are generally developed either as a WiFi-based WDMX or a ZigBee-based WDMX, but recently, a dual method WDMX device has been developed that utilizes two or more communication methods for increased reliability [3][4]. However, for WDMX devices that transfer lighting-control data in real-time, the interference from surrounding wireless devices hinders the stability of the communication. In particular, the interference between the wireless devices and the data delays cannot guarantee real-time control over the transfer of data. This problem is especially acute at venues such as concerts, where a large crowd has a variety of wireless devices with them. In such environments, real-time control of data transfers is even more difficult to guarantee. Therefore, this study has developed a novel concept: a WDMX device featuring the ability to load data that has been stored in advance and transmit it to the lighting device in real-time. The proposed device saves predesigned light-production data and various effects in the receiving WDMX device before the performance, using real-time lightingcontrol data transmissions. Thus, it sends only simple command signals during the actual performance 2 DMX512 PROTOCOL DMX is a communication method for transmitting digital control signals using a twisted pair cable. The US Institute for Theatre Technology (USITT) developed the DMX protocol in 1986 to control dimmers. The DMX512 protocol is an improved version of DMX, developed in 1990, and has gradually been formally adopted by manufacturers. DMX512-based control devices have been used to create lighting effects and for dimming,

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 1. A DMX512 Frame Format.

along with special lighting fixtures for stages and theaters [5]. Figure 1 shows the DMX512 frame format according to which a DMX512 frame is transmitted based on RS-485. The DMX512 protocol is an 8-bit asynchronous serialcommunication method that operates at a rate of 250 kbps and has a pulse period of 4 us. A single data slot is 11 bits in size, and uses 8 bits for the data, 1 bit for the start bit, and 2 bits for the stop bit. A DMX512 controller has 512 channels (or slots) for one connection path, and sends a value represented by 1 byte to the individual channels. In other words, up to 512 channels can be controlled and individual devices are managed with a resolution of 8 bits. 3 PROPOSED WDMX SYSTEM The WDMX proposed in this study is not designed to send control or effect data in realtime, during the performance. Rather, it is designed to send the data to the receiver-side of the WDMX before the performance, where it can be stored. Therefore, when the pre-stored show number or effect number is sent wirelessly to the WDMX receiver through the WDMX transmitter from the main controller during the performance at the time the effect is needed, the WDMX receiver loads the prestored production data and sends that data to the stage controller (e.g., to the dimmer). For instance, for a set of ten songs, the director might create ten light performances that are synchronized with the songs. With the proposed method, these can be designed in advance using the stage lighting and equipment. With the proposed system, the production data from the pre-produced light performances is stored

154


Figure 2. A Proposed Wireless Lighting Control System.

wirelessly in the WDMX receivers that are using the WDMX transmitter on the main controller before the start of the performance attached to each stage controller (e.g., dimmers). Consequently, it is easier to apply changes to the performance production, even when all the equipment is installed before the start of the show. During the actual performance, the preproduced light show is managed by wirelessly sending simple command data (including the number and play-time information for each song) to the DMX receiver from the main controller at the beginning of each song in accordance with the performance. Thus, when the play signal for pre-stored effects with various external inputs is wirelessly sent from the main controller to the DMX receiver, the stored effects are sent from the DMX receiver to the stage-production controller, triggering the appropriate effect for the external input to perform in real-time, and thus, creating an interactive performance. The WDMX proposed in this study was also configured to operate in real-time data-transfer mode in order to be compatible with existing systems. The data-recording mode and the dataplayback mode were added to minimize the uncertainty caused by interference with external wireless equipment during a performance. 3.1 System Design Figure 3 shows the transmitter and receiver system for the proposed WDMX. The WizFi250 module by Wiznet was used for

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

(a)

(b) Figure 3. A Schematic Diagram of Proposed WDMX: a) is transmitter and b) is receiver.

wireless communications. The operating mode can be selected externally using hardware, which is a switch on the side of the transmitter that toggles the operating mode. Furthermore, an external switch was placed on the receiverside in order to set the ID for the receiver module, along with a configuration that can save the production data via external memory. RS-485 was used as the hardware interface for both the transmitting and receiving units for compatibility with existing wire-based controllers. For the WDMX transmitter, the signal for the RS-485 level that is sent from the main controller goes through the RS-485 drive and is converted to a transistor-transistor logic (TTL)

155


level-signal that is then sent to the microprocessor. The transmitted data is analyzed in the microprocessor, which uses the First In First Out (FIFO) method. This data comprises the internal SRAM, and is then sent to the WiFi module through a Universal Asynchronous Receiver/Transmitter (UART). For the WDMX receiver, the signals that come in through the wireless module are processed and, depending on the operating mode, they are either saved on an SD card or used to call saved data from SD card as a DMX512 signal and then outputted via the 485 drive. The WDMX transmitter/receiver uses the User Datagram Protocol (UDP) by default, and is designed for a single transmitter to send data to multiple receivers for real-time transmissions and commands. When a command signal or the external switch trigger recording mode, the transmitter and specified receiver use a TCP/IP socket for sending and receiving data that is then stored. 3.2 WDMX Protocol As explained above, the WDMX device proposed in this study aims to secure the versatility of a WDMX device by replacing existing wired transmissions with wireless transmission (real-time relay) and by initiating play in actual performances through command signals using production data sent and saved in advance. To this end, a WDMX protocol was developed for saving and playing performance data. This protocol is distinct from the existing DMX512 protocol. The existing DMX512 protocol was nevertheless retained for real-time sending in order to ensure the system’s versatility. The DMX512 protocol was modified, however, for save-and-play mode, as shown in Figure 4. The start frame of the WDMX signal was modified in order to distinguish it from an existing DMX512 signal. In recording mode, the IDLE signal from the W-DMX signal was set to a minimum of 3 ms to allow sufficient time for the WiFi wireless module to transmit the data at a rate of

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 4. A Proposed W-DMX Protocol.

250 kbps while the packet is being received. This was done in order to prevent the data from overflowing in the transmission unit. Figure 4 illustrates the W-DMX protocol proposed in this study. The W-DMX protocol sets the mode for the trans-receiver with the start frame. When the start frame is 0x00, it is in streaming mode and uses the same protocol as the existing DMX512. A UDP socket is used to send data for real-time data transmissions. When the start frame is 0xaa, it is in recording mode and the production data that is to be saved after the start frame is sent through TCP/IP socket, ensuring high reliability. Finally, when the start frame is 0x55, it is in command mode, allowing for the construction of data in channels 1-4, with data sent using a UDP socket. 4 EXPERIMENTS and RESULTS In order to verify the WDMX system proposed in this study, Arduino—an open-source platform for designing interactive projects— was used for testing. An oscilloscope was used to confirm that the DMX512 data coming in from the console is being transmitted correctly in each mode through the WiFi module and Wizfi250. Furthermore, a dimmer from Feelux was connected to the WDMX transmitter to check the operation of the W-DMX protocol as shown in Figure 5.

156


REFERENCES

Figure 5. A WDMX Test Environments.

[1]

J. T. Tsao, “Solid-state lighting: lamps, chips, and materials for tommorrow,” IEEE Circuits and Device Magazine, vol. 20, no. 3, pp. 28-37, June 2004.

[2]

S. Shon, J. J. Woo, and Y. Han, “Implementation of a novel WDMX controller for LED lights”, Journal of Korean Institute of Illuminating and Electrical Installation Engineers, vol. 22, no. 10, pp. 1-7, 2008.

[3]

H. Byun, C. Feng, and S. Shon, “Implementation of a WLAN DMX Server based on NDIS WLAN miniport driver”, International Journal of Smart Home, vol. 7, no. 4, pp. 361-370, July 2013.

[4]

J. I. Kim, and B. H. Hwang, “Implementation of broadband LED lighting system using ZDMX modules based DMX512 protocol”, Journal of Korea Contents Association, vol. 10, no. 11, pp. 38-47, 2010.

[5]

F. Fubo, L. Youyuan, and C. Dingfang, “The protocol and realization of DMX512”, Journal of China Illuminating Engineering, 2002.

5 CONCLUSIONS In this study, we proposed a novel WDMX lighting-control system that is distinct from existing wired/wireless DMX devices used for large-scale LED light production. The WDMX proposed in this study is a wireless DMX transreceiver that uses WLAN, and improves upon the shortcomings of existing wired transmissions, especially with regard to installation and maintenance at large-scale venues. In particular, in order to strengthen the security during real-time data transmissions, which is rather weak in conventional WDMX trans-receivers, the proposed WDMX offers a save-and-play feature designed to be compatible with existing DMX systems. Furthermore, by presenting a WDMX protocol that is compatible with the existing DMX512, the versatility of the proposed WDMX was ensured. The feasibility of the proposed WDMX system was verified in a laboratory environment, and further testing in a large-scale performance environment is expected to follow. ACKNOWLEDGEMENT

This research is supported by Ministry of Culture, Sports and Tourism (MCST) and Korea Creative Content Agency (KOCCA) in the Culture Technology (CT) Research & Development Program.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

157


A Study on Motion-Based UI for Running Games with Kinect Jimin Kim, Pyeong Oh, Hanho Lee, Sun-Jeong Kim* Interaction Design Graduate School, Hallym University 1 Hallymdaehak-gil, Chuncheon-si, Gangwon-do 200-702 Korea {Rudengkim, vudhrh}@gmail.com, [email protected], [email protected] ABSTRACT This study examines the efficiency of human motion-based UI for video games with motion capture system, Kinect. We took an investigation to play with the Kinect sensor in the running game which was developed and designed using two kinds of UI. One UI consists of more intuitive and familiar motions such as turning and jumping. The other UI consists of arm motions like raising hands. As a result, UI with arm motions was easier for users to master and results in higher success rates to play than the other UI. Therefore we can conclude when a game is developed using Kinect and its UI is configured with motion recognition, the motion with the arms rather than the other parts of the body helps player better to enhance the play skills and immerse in the game.

KEYWORDS NUI (Natural User Interface), Kinect, Video Games, Unity3D, Motion Recognition

1 INTRODUCTION With the development of computers, the input devices for human computer interaction have been diversified into keyboard, mouse, touch pad, speech recognition, and so on. Nowadays many people have actively researched on NUI (Natural User Interface) especially using Kinect which is a motion sensor that came onto the market at low price by Microsoft on 2011. Kinect as gesture recognition sensor is showing potential as an interface of new generation replacing the mouse and keyboard. Some studies show the potential of using motion based interaction for learning. For example

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

‘Touching Notes’ present how gesture based interfaces can stimulate and motivate children into learning the basics of music notation [1]. For motivating students and enhancing effectiveness, ARCS model of motivation design is considered during developing the Kinect sensor-assisted game based learning system [2]. ARCS model consists of four major steps for learners to become and remain motivated in the learning process: Attention, Relevance, Confidence, and Satisfaction. In addition to learning system, Kinect is utilized in the area of rehabilitation. An interactive gamebased rehabilitation tool for balance training of adults with neurological injury was developed [3]. Instead of WiiFit, Kinect provided markerless full-body tracking on a conventional PC. The Kinect-based rehabilitation game, “JewelMine” consists of a set of static balance training exercises which encourage the players to reach out of their base of support [4]. Most previous studies dealt with the accuracy of the recognition of an implementation of a gesture, but the efficiency and convenience of UI itself were not considered. Our research motivation is searching more natural and efficient motions to make it easy for users to play games with Kinect. In this study we focus on analyzing motion-based UI for running games with Kinect. 2 MOTION-BASED UI WITH KINECT We implemented a running game using a game engine, Unity3D (ver. 4.6.0) and Kinect SDK (ver. 1.8). In common running games, a main character is usually running during play. To avoid obstacles, players choose to jump them or

158


turn left or right. For input of jumping and turning, we implemented two kinds of motionbased UIs and then examined which one is more natural and efficient for users to play.

display ten prefabs at each frame, and among them we place five obstacles in the scene. We prepare five types of obstacles like Figure 1.

(a) (a)

(b)

(b) (c)

(d) (c)

(e) Figure 1. Five obstacles are used in our running game.

2.1 Implementation of a Running Game In our game, the road where a main character runs is randomly generated using prefabs. We

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

(d) Figure 2. Screen shots of our running game.

159


Figure 2 shows several screen shots of our running game. A main character keeps running and a user selects to jump or turn left or right whenever she meets an obstacle. If she fails to avoid an obstacle or gets out of the road, the game is over. The game score is the run distance. To obtain human motion data from a Kinect motion sensor in Unity3D application, we used “Kinect Wrapper Package for Unity3D” [5] which is provided by Entertainment Technology Center (ETC) in Carnegie Mellon University. After importing this package we could use assets for motion tracking. To obtain the position data of joints we modified the function of Update() in the script ‘KinectPointController.cs’ which is attached to the prefab ‘KinectPointMan.’ Kinect tracks the skeleton and a tracked skeleton provides the information about the positions of twenty joints of the user’s body (Figure 3).

where an unit vector u is (X1‒Cx, Y1‒Cy) / ||(X1‒Cx, Y1‒Cy)||, an unit vector v is (X2‒Cx, Y2‒Cy) / ||(X2‒Cx, Y2‒Cy)||, and (Cx, Cy), (X1, Y1), and (X2, Y2) are the coordinates of joints’ positions (Figure 4).

Figure 4. A joint angle θ consists of three joints.

As a matter of fact, the position of a joint is 3D coordinates. However we carried out the calculation in 2D for speed. In case of turning gestures, the coordinates of joints’ positions were projected onto XY plane and then a joint angle was computed in 2D. In case of a jumping gesture, the coordinates of joints’ positions were projected onto YZ plane and then a joint angle was computed in 2D too. 2.2 Design of Motion-Based UIs For experiments we designed two kinds of motion-based UIs. One UI was designed with more intuitive and familiar motions like Figure 5. Figure 5 (a) shows a gesture to make a game character to turn right. Figure 5 (b) shows a gesture to make a game character to jump.

Figure 3. Twenty joints’ names which Kinect tracks [6].

To recognize a gesture, we should compute a joint angle θ. We used an equation as follows: θ = acos( u • v )

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

(1)

(a) (b) Figure 5. Motions for turning and jumping.

160


To turn, a user leans the upper body to left or right side. To recognize this motion, we got the positions of three joints: Head, Shoulder Center, and Hip Center. We decided whether to turn left or right by comparing the positions of x coordinates of Head and Shoulder Center. For example, if the value of x coordinates of Head is less than the value of x coordinates of Shoulder Center, this gesture is turning left. Otherwise, it is turning right. To jump, a user bends the knee. To recognize this motion, we got the positions of three joints: Hip, Knee, and Ankle. After computing a joint angle, if this angle is within threshold range, we decided that a user makes a motion for jumping.

Because joint angles of initial standing posture are not equal to zero, we defined the range of threshold for each joint angle (Table 1). Table 1. The range of threshold for joint angles (degrees) Types of Initial Maximum Minimum Motions Value Value Value UI with natural motions UI with arm motions

Turning

-10~10

60

180

Jumping

0~15

60

180

Turning

45~55

20

140

Jumping

45~57

-90

-30

3 EXPERIMENTS and RESULTS To compare two types of motion-based UIs, eleven university students took part in an experimental investigation. For each UI they played our running game ten times. We averaged top five game scores per person. Figure 7 shows its result. In the graph, vertical values (0~120) represent the average of game scores and horizontal values (1~11) represent the ID of person.

(a) (b) Figure 6. An arm motions for turning and jumping.

The other UI was designed with arm motions like Figure 6. This UI design focuses on the convenience of gestures. Figure 6 (a) shows an arm gesture to make a game character to turn right. Figure 6 (b) shows an arm gesture to make a game character to jump. To turn right, a user raises the right hand. To recognize this motion, we got the positions of three joints: Elbow Right, Shoulder Right, and Hip Right. To detect the motion of turning left, we get the position of three joints such as Elbow Left, Shoulder Left, and Hip Left. To jump, a user raises both hands. To recognize this motion, we got the positions of six joints: Elbow Left, Elbow Right, Shoulder Left, Shoulder Right, Hip Left, and Hip Right.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Figure 7. The average of game scores for eleven test subjects.

As a result, UI with arm motions is easier for users to master and results in higher success rates to play our running game than the other UI. We become aware of its reason as follows: people frequently use arms to express their judgment rather than other parts of the body

161


and therefore arm motion becomes to be easier for users and trained more precisely than the motion of other parts of body.

In proceedings of VRW 2012 (IEEE Virtual Reality Short Papers and Posters), pp. 171-172, March 4-8, 2012. [5]

http://wiki.etc.cmu.edu/unity3d/index.php/Microsoft _Kinect_-_Microsoft_SDK

[6]

https://msdn.microsoft.com/enus/library/jj131025.aspx

4 CONCLUSIONS In this study we designed two types of motionbased UI for a running game with Kinect. One UI consists of more natural and intuitive motions such as leaning the upper body to left or right side for turning left or right. The other UI consists of only arm motions like raising both hands for jumping. As the result of experiments, game players prefer arm motions because people usually use their arms to express their decision in daily life and it makes users easy to enhance the play skills, which helps them to immerse in the game. Therefore, for improving the mastery and immersion level of players, it is better to design motion-based UI with arm motions rather than the motion of other parts of body. ACKNOWLEDGEMENTS This research was supported by Hallym University Research Fund (HRF-201409-012). REFERENCES [1]

M. Renzi, S. Vassos, T. Catarci, and S. Kimani, “Touching Notes: A Gesture-Based Game for Teachingth Music to Childer,” in proceedings of TEI 2015 (9 International Conference on Tangible, Embedded and Embodied Interaction), Stanford, CA, USA, pp. 603-606, January 15-19, 2015.

[2]

C.-H. Tsai, Y.-H. Kuo, K.-C. Chu, and J.-C. Yen, “Development and Evaluation of Game-based Learning System Using the Microsoft Kinect Sensor”, International Journal of Distributed Sensor Network, in press.

[3]

B.Lange, C.-Y. Chang, E. Suma, B. Newman, A.S. Rizzo, and M. Bolas, “Devlopement and evaluation of low cost game-based balance rehabilitation tool using the microsoft kinect sensor”, In proceedings of EMBC 2011 (IEEE Engineering in Medicine and Biology Society), pp. 1831-1834, August 30September 3, 2011.

[4]

B. Lange, S. Koening, E. McConnell, C. Chang, R. Juang, E. Suma, M. Bolas, and A. Rizzo, “Interative game-based rehabilitation using Microsoft Kinect”,

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

162


Evolution of Multimodulus Algorithm Blind Equalization Based on Recursive Least Square Algorithm Sardar Ameer Akram Khan and Shahzad Amin Sheikh Electrical Department, College of Electrical and Mechanical Engineering , NUST Pakistan E-mails: [email protected] , [email protected]

ABSTRACT Blind equalization is an important technique amongst equalization family. A Multimodulus algorithm based on blind equalization removes the undesirable effects of ISI and cater ups the phase issues, saving the cost of rotator at the receiver end. In this paper a new algorithm combination of recursive least square and Multimodulus algorithm named as RLSMMA is proposed by providing few assumption, fast convergence and minimum Mean Square Error (MSE) is achieved. Excellence of this technique is shown in the simulations presenting MSE plots and the resulting filter results.

KEYWORDS: Blind Equalizations, Constant Modulus Algorithm, Multimodulus Algorithm, Recursive Least square Algorithm, Quadrature Amplitude Modulation (QAM).

1 INTRODUCTION Equalization is the basic building block of modern digital communication system. Equalization is defined as the process which helps to retrieve the transmitted signal free from the undesirable channel effects at the receiver end. Undesirable effect can be classified as (linear) channel distortion and additive noise commonly known as (ISI) that corrupts the transmitted signal making it cumbersome for the receiver end to recover the transmitted data directly.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Equalization can be categorized into two major categories the Non-Blind equalization and the Blind equalization. Non blind equalization is a technique which equalizes the received signal by the help of training bit to update the weights. Training bits are embedded with the transmitted signal and repeated every time, information about the training bits are pre known at the receiver so thereby the received training bits are analyzed and the channel response is calculated accordingly to match the equalizer output minimizing some criterion typically MSE(Mean Square Error). Non blind equalization gives a better performance but the main disadvantage is the wastage of bandwidth about 25% bandwidth consumption in GSM (global system for mobile communication) [1]. Blind Equalization or self-recovering algorithm defines an equalization without the use of training bits [3], [9]. Blind equalization has been a hot area of research from the last few decades. Constant Modulus Algorithm is considered as one of the famous algorithms of blind equalization proposed by Godard [1], [2], [3]. CMA utilizes method of steepest descent to equalize the signal. CMA due to its increased bandwidth efficiency which increases bits rate [1], and its simplicity like

163


LMS makes it very popular but its major weakness is that it has slow convergence seemingly its cost function is also dependent on the amplitude of the signal thus resulting lack of knowledge about the constellation, hence the overall performance suffers when using higher order QAM schemes [4], [5]. A replacement of CMA was proposed called as MMA [6], [7]. MMA despite of just minimizing the magnitude of equalizer’s output y(n), it considers both the real yk(n) and yi(n) individually[8]. MMA achieves better convergence and the cost of rotator at the receiver end is also nullified. In this research work a new MMA based algorithm has been proposed by having some enhancement to the Multimodulus algorithm results are quite considerable for the QAM constellations. In section 2 the existing algorithms are discussed which contains the brief description about CMA, RLSCMA and MMA. In section 3 new algorithm is proposed, followed by the simulation comparison included in section 4 and in section 5 the final work is concluded. 2 THE EXISTING BLIND ALGOTHMS 2.1 Constant Modulus Algorithm Constant Modulus Algorithm was basically proposed by Godard [1], [2], [3]. CMA is also one of most famous algorithm amongst the Bussgang algorithm it utilizes memory less nonlinear function in filters output to obtain the results [1]. CMA algorithm is derived from the method of steepest descent when no training bits are utilized [1]. Cost function is formulated by:

where yn is the filter’s output and Rp is the non-negative constant. Cost function is differentiated to get alike LMS algorithm, further it is only dependent on the modulus of filters output so the information about the carrier phase is lost. Equations of CMA algorithm is concluded as:

wi(n+1)=wi(n) + µx(n-i)yn(Rp- |yn|2)

(2)

Step size is denoted by µ, wi(n) is the ith tap equation at a certain time n, similarly x(n) is the input of filter at time n, Rp is positive constant. Error equation is calculated as: e(n)=yn(Rp-|yn|2)

(3)

where e(n) is the error term equation, yn is the equalizer output. Rp is calculated as: (4)

where

is the transmitted source signal.

Constant Modulus Algorithm is preferred over other Godard algorithm due to its promising performance with respect to Mean Square Error (MSE), but due to slow convergence rate and lacking the knowledge about the carrier phase it leads to the research of other algorithms.

(1)

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

164


2.2 Recursive Least Square Constant Modulus Algorithm Due to slow convergence and adaptation of CMA algorithm, Nassar Amin and Nahal Waleed introduced a modification of CMA algorithm and used Method of Least square over Method of steepest descent algorithm for CMA [11]. RLSCMA outperform CMA in term of MSE [11], [12]. Cost function of RLSCMA for the fast convergence is specified by: J(w)=E[(|y(n)|2 - 1)2]

(5)

J(w)=∑

(6)

(|y(n)|2-1)2

is the forgetting factor and valued between. . RLSCMA is derived from standard Recursive least square algorithm except it considers an input signal z(n) [11], rest of algorithm is concluded as: z(n)=x(n)xH(n)w(n-1)

(7)

h(n)=p(n-1) * z(n)

(8)

p(n-1) is the inverse correlation matrix and Kalman gain is calculated as follow: k(n)= e(n)=1 - wH(n-1)z(n)

(9) (10)

e(n) is the error term equation. Weight update equation is as follow: w(n)=w(n-1)+k(n)e*(n)

(11)

P(n) inverse correlation matrix update is as follow: (12)

initial conditions are specified as w(0)=[1,1x(k-1)] ,p(0)=ẟ-1Ikxk delta ẟ is assumed small positive constant like 10-3.

2.3 Multimodulus Algorithm Multimodulus algorithm (MMA) was introduced by Yang et al. [6], [7]. Carrier phase deficiency of CMA was catered in the this algorithm the cost function equation of MMA is as follow: JMMA(n)= JR(n)+ JI(n)

JMMA(n)=E[(y2R(n) -R2,R)2]+ E[(y2I(n) -R2,I)2] (14)

where yI(n) and yR(n) are the imaginary and real terms of equalizer’s output. R2,R and R2,I can be calculated as per following equations: R2,R=

(15)

R2,I=

(16)

where sI(n) and sR(n) are the imaginary and real part of the transmitted source signal. Weight update equation is given as: w(n+1)= w(n) - µe(n).x*(n)

(17)

similarly the error term is also the sum of real part of error eR(n) and the imaginary counterpart of error eI(n) given in following equations: eR(n)= yR(n)( yR2(n)eI(n)= yI(n)( yI2(n)e(n)= eR(n)+ eI(n)

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

(13)

) )

(18) (19) (20)

165


MMA can remove ISI and explicitly resolves the carrier phase issues due to its better phase tracking capability as stated above. However MMA does not performs well for the dense constellations, so to cater up this issue a new algorithm is proposed for fast convergence and providing better results for dense QAM.

p(n-1) is defined as inverse correlation matrix. Kalman gain is as follow: k(n)= eR(n) = yR(n)( yR2(n)eI(n) = yI(n)( yI2(n))

(25) )

(26) (27)

e(n)= eR(n)+ eI(n)

(28)

R2,R=

(29)

R2,I=

(30)

3 NEW BLIND ALGORITHM A new blind algorithm Recursive Least Square Multimodulus Algorithm (RLSMMA) is proposed in this section. Combining the work done by S.Makino and Y.Kaneda for Recursive Least Square [13], [14], [15], [16]. Chen and et.al and for Recursive Least Square Constant Modulus Algorithm [17], [18] and various links for CMA and MMA convergence and MSE optimization a new algorithm is developed. Performance of the equalizers is analyzed by the MSE plots. Cost function of the algorithm is defined as: J(w)=E[((y2R(k)-1)2]+E[((y2I(k)-1)2] (21) ((y2I(k)-1)2 J(w)=∑ (y2R(k)-1)2+ ∑ (22)

where yI(k) and yR(k) are the imaginary and real terms of equalizer’s output and is the forgetting factor valued between. .

Algorithm of RLS-MMA is as follow, z(n) is the input of the equalizer: z(n) = x(n)xH(n)w(n-1)

(23)

h(n) = p(n-1)z(n)

(24)

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

where sI(n) and sR(n) are the imaginary and real part of the transmitted source signal.

Weight update equation is given as: w(n)=w(n-1)+k(n)e*(n)

(31)

P(n) inverse correlation matrix update is as follow: (32)

4 SIMULATION RESULTS The following section evaluates the performance of the entire algorithm mentioned where the main parameter focused was lay on the Mean Square Error (MSE). Performance of CMA and RLSCMA was compared further MMA performance was compared with the proposed algorithm RLSMMA. In the simulations s(n) was the QAM signal, SNR is 30db. All the simulation considered a channel of a seventh order FIR filter similarly the equalizer is

166


also FIR adaptive equalizer of seventh order is considered. CMA and MMA simulations used a stepsize µ of value 0.09 while forgetting factor is considered 0.99 for RLSCMA and RLSMMA.

Results are shown as follow:

Figure 3. Equalized results of MMA

Figure1. Comparison of CMA and RLSCMA

Figure 4. Equalized results of RLSMMA

Figure 2. Comparison of MMA and RLSMMA

Figure 5. Error results of RLSMMA

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

167


Figure 1 shows the performance comparison of CMA and RLSCMA. RLSCMA outperforms CMA as it attains fast convergence and Mean Square Error is also lower then CMA. Figure 2 shows the performance analysis of MMA and RLSMMA. It is clear that the performance of the newly proposed RLSMMA is much better than MMA for dense QAM as MSE is much lower in the RLSMMA case then MMA, clearly indicating the promising results as shown in figure 3. Figure 5 shows the error results of the proposed new blind equalizer. The results are comparable and quite promising.

5 CONCLUSION A new blind equalization technique RLSMMA was introduced in this paper. Simulations were carried on 16QAM symbol sequence. Results obtained for RLSMMA were quite promising as compared to the conventional blind equalization algorithms. The main advantages over other can be evaluated as follow, high rate of adaptation, minimum mean square error, less computational complexity, cost of rotator at receiver end saved. References [1]

Godard Dominique, “Self-Recovering Equalization and Carrier Tracking in TwoDimensional Data Communication Systems", IEEE Transactions, vol. 28, no. 11, pp.18671875, 1980.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

[2] [3] [4]

[5]

[6]

[7]

[8]

[9]

[10] [11]

[12]

[13]

Proakis John, “Digital Communications” Columbus McGraw-Hill Companies 2001. Haykin Simon, “Adaptive filter theory”, Pearson Education India, 2008. Giannakis Georgios , Steven Halford, "Blind Fractionally Spaced Equalization of Noisy FIR Channels Direct and Adaptive Solutions", Signal Processing, IEEE Transactions, vol. 45, no. 9, pp.2277-2292, 1997. Gesbert David, Pierre Duhamel, Sylvie Mayrargue, "On-line Blind Multichannel Equalization Based on Mutually Referenced Filters", Signal Processing, IEEE Transactions, vol. 45, no. 9, pp. 2307-2317, 1997. Yang Jian, Jean-Jacques Werner, Dumont, "The Multimodulus Blind Equalization Algorithm", Digital Signal Processing Proceedings, 13th International IEEE Conference, vol. 1, pp.127-130, 1997. Yang Jian, Jean-Jacques Werner, Dumont, "The Multimodulus Blind Equalization and its Generalized Algorithms", IEEE Journal on Selected Areas in Communications, vol. 20, no. 5, pp.997-1015, 2002. Shahzad Amin Sheikh, Pingzhi Fan, "A New Multimodulus Blind Equalizer for Dense QAM Constellations", Wireless Mobile and Multimedia Networks, 2006 IET International Conference, vol. 1, no. 4 , pp.1-4, November 2006. Proakis John, Dimitris, Manolakis, “Introduction to Digital Signal Processing”, Prentice Hall Professional Technical Reference, 1988. Ding Zhi, Ye Li, “Blind Equalization and Identification”, CRC press, 2001. Nassar Amin Mohamed, Nahal Waleed, "New Blind Equalization Technique for Constant Modulus Algorithm (CMA)", Communications Quality and Reliability (CQR), 2010 IEEE International Workshop Technical Committee, vol. 1, no. 6, pp.1-6, June 2010. Cioffi John, Thomas Kailath, “Fast RLS Transversal Filter for Adaptive Filtering”, IEEE Transaction on ASSP, vol. 32, no. 2, pp.407-410, April 1984. Shoji Makino, Yutaka Kaneda, Koizumi, "Exponentially Weighted Stepsize NLMS Adaptive Filter Based on the Statistics of a Room Impulse Response", Speech and Audio

168


[14]

[15]

[16]

[17]

[18]

[19]

[20]

Processing, IEEE Transactions vol. 1, no. 1, pp.101-108, January 1993. Shoji Makino, Yutaka Kaneda, "A New RLS Algorithm Based on the Variation Characteristics of a Room Impulse Response" Acoustics, Speech, and Signal Processing, ICASSP 1994, IEEE International Conference, vol. 3, pp.19-22, 1994. Shoji Makino, Yutaka Kaneda, “ES-RLS (Exponentially Weighted Stepsize RLS) Algorithm Based on the Statics of a Room Impulse Response”, proceeding on Acoustics, Japan, pp.547-584, October 1992. Makino Shoji, Yutaka Kaneda, "Exponentially Weighted Stepsize Projection Algorithm for Acoustic Echo Cancellers", IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences, vol. 75, no. 11, pp.150-151, 1992. Y. X Chen, "RLS Adaptive Blind Beam forming Algorithm for Cyclostationary Signals", Electronics Letters, vol. 35, no. 14, pp.1136-1138, 1999. Chen, Yuxin, "Recursive Least Squares Constant Modulus Algorithm for Blind Adaptive Array", Signal Processing, IEEE Transactions, vol. 52, no. 5, pp.1452-1456, 2004. Ali Sayed, Thomas Kailath, "A State-Space Approach to Adaptive RLS Filtering", Signal Processing Magazine IEEE , vol. 11, no. 3, pp.18-60, 1994. Simon Haykin, “Adaptive Tracking of Linear Time-Variant Systems by Extended RLS Algorithms", Signal Processing, IEEE Transactions, vol. 45, no. 5, pp.1118-1128, 1997.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

169


Simple Heuristics for the Choquet Integral Classifier Ken Adams Sino-British College, University of Science and Technology 1195 Fuxing Middle Road, Shanghai China, PRC [email protected]

A data set, with one decision attribute,

Abstract

consists of a finite number of condition The

Choquet

Integral

is

a

successful

classification method. However, like other methods, when applied to large data sets where many coefficients have to be optimised,

attributes (variables) X  x1, x2 , , , , xn  and a condition attribute y that can have a number of values. A record is one observation of all

search methods such as genetic algorithms

the attributes. The value of an attribute xi

(GA) are used. In this paper, heuristics are

over all records is considered to be a function

developed

of

that

can

achieve

reasonably

accurate results by using information gained from the data set. This enables optimisation using a smaller set of coefficients than those needed in the original search space and this may provide a good starting place for a GA search and other optimisation methods. For the purposes of this research, the data used is

xi

and

it

is

customary

to

write

f j ,i  f j  xi  as the j th observation of the ith

attribute, [1]. The classification occurs when information from existing records is used to predict the value of the decision attribute for a new record when only its condition attributes are known.

the Wisconsin Breast Cancer data set, and it is envisaged that the methods described here will generalise onwards to many other optimisation techniques and data sets.

With many variables to explore, search algorithms such as genetic algorithms (GA) are often used. The idea of genetic algorithms

Key Words: Data mining, Optimisation,

was invented by Holland [2] Looking at

Genetic Algorithm, GA, Choquet Integral,

nature Holland considered Darwin's theory of

Heuristics, Classification, Wisconsin Breast

natural selection, i.e. that the fittest members

Cancer.

of the species survived to breed more often

1 INTRODUCTION

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

and thus produced more offspring. As offspring inherit their characteristics from

170


their parents, fit members of a population

and generate a simpler set of rules than the

would have a better chance of passing their

original. The proposed activity heuristic is

characteristics on to the next generation than

shown to produce better results than a

unfit members. Over generations the species

previously-published rough set analysis of the

would evolve as the average fitness of its

same data; and shows comparable results to

members rises. It is also possible that some

the well-known ID3 information theoretic

random change might appear in a member of

approach.

the next generation. If this random change is of benefit then that individual will be successful and the random change will be passed on to yet another generation.

The Choquet integral is a well-known classification method. It requires that a coefficient, called a measure be allocated to every subset of X the condition attributes. In

Heuristics are often used to direct the search

this paper heuristics are developed to find a

in a possibly useful direction that may achieve

good set of coefficients. Genetic algorithms

a result of high fitness by the GA [3, 4]. A

are a common tool for finding the large

heuristic can be considered as some simple

number of coefficients needed to find the

and fast technique which, although may not

measures needed for Choquet classification

find the optimum will often find a good place

[1], [9]. It is anticipated that fresh ideas on

to start. One or more of the members of the

heuristics will help future authors with GA

population are heuristically enhanced before

and other searches.

the search begins. Blind heuristics, for example bit-flipping [5, 6] can be useful but use no information gathered from the data. In Adams [7] Chapter 6 heuristics based on the autocorrelation and on the discrete Fourier transform were developed. They were used to

The data used in this experiment is the wellknown Wisconsin Breast Cancer data set [10]. Results reported in this paper are comparable to others reported for the same data set in the literature.

optimise the number of coefficients of

Sections to follow are: §2 Wisconsin Breast

multiple-valued polynomials. Results were

Cancer Data Set, §3 Description of the

shown to be significantly better than running

discrete Choquet integral §4 Rationale (of the

the genetic algorithm without heuristics.

heuristics)

Heuristics methods can be a successful

§5

Heuristics

Developed

§6

Results and Comparisons.

technique in their own right. In [8] a novel

2. WISCONSON BREAST CANCER

heuristic (activity of the variables) was used

DATA SET

to partition a decision table into sub-tables

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

171


The Wisconsin Breast Cancer data set is

10. Mitoses: 1 - 10

available at [10]. It consists of 699 records of

11. Class: (2 for benign, 4 for malignant)

patients who were examined for breast cancer and had various measurements taken. As sixteen of the records have a missing value

3 DISCRIPTION OF THE DISCRETE CHOQUET INTEGRAL

only the 683 complete records are used here,

The Choquet integral was proposed by

of these 239 had cancer and 444 had not. The

Choquet [13] to study capacities in the field

data was collected from January 1989 until

of economics. The discrete Choquet integral

November 1991 by Dr. WIlliam H. Wolberg

is used as an aggregation technique and has

(physician) at the University of Wisconsin

been

Hospitals Madison, Wisconsin, USA [11, 12].

classification purposes, including [14] and

The scale of each measurement is an integer

developed

by

many

authors

for

[15]. The integral takes the form:

in the range one to 10, and the attribute information for the data is given below. For this paper the first condition attribute x1 is

C  f

n

     

d   f xi*  f xi*1   xi* , xi*1 ,...x n*



i 1

(1)

Clump Thickness and the last condition attribute x9 is Mitoses. The decision attribute

Where the variables xi* are a permutation of

y is Class. In order to integrate with the

the x i into ascending order of their value.

authors existing software, the decision class

Here  is a measure or weighting associated

was relabelled as y  1 meaning benign and

with each subset.

y  2 meaning malignant. Attribute Information: 1. Sample code number: id number 2. Clump Thickness: 1 - 10 3. Uniformity of Cell Size: 1 - 10 4. Uniformity of Cell Shape: 1 - 10 5. Marginal Adhesion: 1 - 10 6. Single Epithelial Cell Size: 1 - 10

A three variable example of how the integral is calculated will now be given. Suppose

X  x1, x2 , x3  , then there is a lattice of subsets [see Figure 1]. For three variables there are eight subsets in the lattice. The lattice is partially ordered by inclusion. A weighting  called a measure is allocated to each subset. In this example the weights are

7. Bare Nuclei: 1 - 10

as follows: {}  0 , {x1}  4 , {x2}  8 ,

8. Bland Chromatin: 1 - 10

{x3}  2 , {x1, x2}  0.4 {x1, x3}  0.5 ,

9. Normal Nucleoli: 1 - 10

{x2 , x3}  0.6 , {x1, x2 , x3}  1 .

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

172


contribution each condition attribute makes to the information available to estimate the decision attribute. However, the Choquet Integral also aims to take account of the interaction between variables, and has many applications such as for optimisation [1], to shift work [16], for multiple regression [17], and for decision rules [18]. There follows an example of classification Figure 1 Lattice of Sub-sets

using the Choquet Integral on a data record

Now suppose that a record has the following

from the Wisconsin data set [Note: It is

values f x1   4 , f x2   8 , f x3   2 . Firstly

unnecessary to calculate differences when

the variables are re-arranged in ascending

adjacent attributes in ascending order have the

order of size of their values, x3 , x1, x2 . Then

same value]. The nine condition attributes are

starting from the top at the universal set X , a

labelled from x1 to x 9 and their values

path is made through the lattice. At each step

recorded in Table 1:

the variable with the next lowest value drops out. {x1, x2 , x3} → {x1, x2} → {x2} → {} . The measure for each subset in the path is

Table 1: Example Record Att.

x1

x2

x3

x4

x5

x6

x7

x8

x9

F(x)

6

8

8

1

3

4

3

7

1

multiplied by the difference between the lowest value of the variables in the subset and the value of the variable that has just dropped

After sorting in ascending order the table

out. Then all the products are added up. The

becomes:

calculations are:

2  01  4 1 0.4  8  4 0.1  3.6 . Aggregation is a way to replace all the

Table 2: Sorted Att.

x4

x9

x5

x7

x6

x1

x8

x2

x3

F(x)

1

1

3

3

4

6

7

8

8

numbers x1, x2 , , , xn by just one number (a sort of averaging). Then if the aggregated figure falls below a certain cut-off the record

Table 3: Calculation

is classified in one way and if above it is

x1 to x9

classified in another. The use of a simple

Att.

weighted average takes into account the

111 111 111

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Diff.



Diff× 

1

1.00000

1.000000

173


111 011 110

2

0.670496

1.340992

through a subset which is close to this global

111 001 010

1

0.656327

0.656327

ratio then there is little evidence either way. If

111 000 010

2

0.656327

1.312654

011 000 010

1

0.300222

0.302222

011 000 000

1

0.302222

0.302222

far from the global probability then passing through the subset may well provide a strong indication of the patient’s condition. Thus

000 000 000 Total

4.914417

expected frequency is the tool used in this paper.

As you can see the value of the integral is

The other key idea is that instead of looking

heavily

measures

for values for all 512 measures (coefficients)

assigned to each subset. There are 512 subsets

and searching through a vast number of

of the nine condition variables of the

combinations of settings. A small set of

Wisconsin data set. It is difficult to allocate

parameters is instead examined. Suppose a

optimum values to all of these coefficients

black box needs to receive k parameters in

thus search techniques like genetic algorithms

order to operate and its job is to generate a

are often used for this and for similar

much larger set of m parameters, needed for

problems.

optimisation. Then it is easier to search the k

dependent

upon

the

parameters than the m directly. This is useful

4. RATIONALE

to get an approximate solution that can then

The data itself provides information on how

be passed through some further optimisation

new data may be processed. If a subset is

procedure. It is believed that this idea can be

traversed by a new record and all the evidence

applied more generally to many other

from the previous data says that any patient

optimisation problems.

whose records passed through that subset has cancer then there is strong evidence to suggest that this patient may have cancer. Likewise, if all information from prior records showed no cancer, it would be expected that the patient’s condition

was

benign.

The

frequency

considered here to be important is the frequency of class two (malignant). Looking at the whole data set of 683 records there are 239 which are malignant. That is an overall probability of 239/683. If a patient passes

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

5 HEURISTICS DEVELOPED The ideas behind the various heuristics are explained in the sub-sections following. The heuristics design a function and if this function is graphed then the horizontal axis (x) is the frequency of class two records passing through that particular subset in the lattice; and the vertical axis (y) is in the range [─1, 1].

174


The expected number of Class2 (cancer

5.1 Mean This simply calculates the mean of the nine condition attributes and then chooses the best cut off point. Using the Choquet integral the

patients) that pass through the subset when the grand totals are used for calculation are:

mean is the result of calculations when the M where M is measure of a subset M is 9

Class1 freq  Class2 freq

TotalTwo TotalOne  TotalTwo

(3)

the cardinality of M [15].

When the frequency of class two

5.2 Class Frequency Formula

(cancer) is exactly that which is For each subset in the lattice, a frequency

expected the ratio calculates to zero.

count is made of the number of records that

The measure is calculated from the ratio

take a path through the subset that are of

in the following manner measure=

either class one (no cancer) and class two

0.5*(ratio +1) this re-scales so that the

(malignant). The grand totals across all of the

range of the measure is the interval [0,

data set are also used. If both frequencies are

1].

zero then the measure is assigned to be zero. When at least one of the frequencies in non-

5.3 Simple Straight line Method

zero the following ratio is calculated:

The measure is calculated using a straight line

Class2 freq  TotalOne  Class1 freq  TotalTwo Class2 freq  TotalOne  Class1 freq  TotalTwo

(2)

that passes through the origin and the point (Class1freq + Class2freq, 1). Here, the horizontal axis is the class two frequency, so

Here TotalTwo is the total number of

if every record is class two then the value one

records

will be calculated. If there are no class two

with

classification

two

(malignant) in the data set; Class2 freq

records, then the measure will be zero.

is the number of records classified as

5.4 Two Straight line Segments

two that pass through this particular subset.

This method is designed that if the frequency of class two is the same as expected, then the

This ratio has the following properties:

y-coordinate will be 0; If all are class two,

If Class2 is zero the ratio becomes  1 .

then the y-coordinate is 1; and if no class two, it is ̶ 1.

If Class1 is zero the ratio becomes 1.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

175


is 0.5  7  20  13.5 and this is to have a y-

5.5 Four Straight line Segments This method is similar to the two straight line method and is designed that if the frequency of class two is the same as expected, then the

value of 0.2. Therefore, there is a change of line segments at the coordinates (13.5, 0.2). Similarly

the

lower

mid-point

is

y-coordinate will be 0; If all are class two,

0.5  0  7  3.5 and this is to have a

then the y-coordinate is 1; and if no class two,

measure of  0.3 . The graph shows the

it is ̶ 1.

example function calculations with all four line segments

However instead of just one straight line between the expected value and the sum of the frequencies two straight lines are used. The mid-point is used as an additional xcoordinate (frequency) and the y-coordinate can be set arbitrarily. The two lines will intersect at these coordinates.

A search

process can be used to find a suitable ycoordinate.

Figure

2:

Example

with

Four

Line

Segments

Similarly the mid-point between the origin and the expected value can be used to generate another pair of lines. The motivation is that frequencies close to the expected value do not provide strong evidence so this provides a way of making their measure even smaller than it would be if a single straight line was used. In total there are four line segments. Obviously more than four line

6 RESULTS & COMPARISIONS A reclassification of the entire data set was made. (Reclassification was performed on an artificial data set in [19] as a proof of concept of the Choquet integral classifier.) The reclassification results for this paper are presented in Table 4.

segments may be used. Here is an example where six records belonging to class one and fourteen belonging to class two, passes through a certain subset. The expected frequency for class two is then

20 

239  6.9985  7 . The upper mid-point 683

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

176


Table 4 Method

Reclassification

Formula

98.22%

One Segment

98.22%

The mean method did well considering its simplicity with a reclassification rate of 97.53%. This result and the closeness of all the various results reported in the literature would suggest that the Wisconsin data set does not stress the classifications techniques

Four Segments

97.88%

hard enough to produce strong differences.

Mean

97.53%

Three papers [20, 21, 22] used only the 683

Two Segments

97.17%

complete records. In the papers the first 400 records were used for training and the last 283 for testing. The resulting testing accuracy’s

The heuristics demonstrated here have shown to be comparable to and often better than the

were:

98.10%,

97.50%

and

98.10%

respectively.

more complicated ideas of neural network and

A comparison of the efficiency of five

decision trees. This fundamentally different

different

and fresh approach should now be explored

Wisconsin Breast Cancer data set was

further on a wider set of benchmarks. The

reported in [23]. The method used was 10-

formula method produced one of the best

fold cross validation and a t-test. The Weka

results with a reclassification rate of 98.22%.

(Waikato

This is satisfying because the structure of the

Analysis) tool [24], which is a downloadable

formula has a logic to it. That if the class two

platform housing a variety of classification

frequency is what would be expected from the

algorithms, was used for the calculations. The

global frequencies then it calculates to zero;

five methods were: Bayesian Network, Naïve

but if the frequency of class two is zero the

Bayes, Multi-layer Neural Network, the

formula returns  1 , and if instead the

ADTree decision tree, and the J4.8 decision

frequency of class one is zero it calculates to

tree. The resulting efficiencies for each

+1. It may be worth considering using a bias

classifier were: 97.20%, 96.11%, 95.58%,

where an extra parameter is added to the

95.49%, and 94.92% respectively.

expected frequency before calculation so that a number higher than (or lower than) the expected frequency is actually the zero. Such a parameter can be global or can even be

classification

Environment

methods

for

on

the

Knowledge

Ten-fold cross validation was performed on four of the methods of this paper results are in Table 5.

optimized for each subset.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

177


Table 5

REFERENCES

Method

10-fold cross validation

Mean

97.51%

Two line segments

96.34%

Formula

96.05%

One line segment

93.27%

[1]

M. Spilde, and Z. Wang, “Solving nonlinear optimisation problems based on generalised Choquet integrals by using soft computing techniques”, Proc. IFSA 2005, pp. 450-454.

[2]

J.H. Holland, (1975), Adaptation in Natural and

As some of the test data would pass through

Artificial

Systems:

University

of

Michigan Press, 1975. [3]

D.E. Goldberg, (1989), Genetic Algorithms in

subsets of the lattice that were not used in the

Search, Optimisation and Machine Learning:

training data, when both the frequencies of

Addision-Wesley Publishing Company Inc., 1989.

class one and of class two were zero the measure was set to M 9 . Methods such as

[4]

Z. Michalwicz, D.B. Fogel, How to Solve It Modern Heuristics: Springer-Verlag, New

those used here are probably more suited for

York, 2000.

large benchmarks were lots of information regarding frequencies can be calculated.

[5]

R. Drechsler, Evolutionary Algorithms for VLSI CAD: Kluwer Academic Publisher, 1998.

The coefficient used for the mean method.

[6]

R. Drechsler, B. Backe, and N. Drechsler,

The mean did slightly better than the best of

(2000), “Genetic algorithm for minimisation

the five methods reported in [23]. The two

of fixed polarity Reed-Muller expressions,”

line segment method and the Formula method

IEE

outperformed

the

lowest

Proceedings

Computers

Digit.

Techniques, Vol. 147, No. 5, Sept. 2000, pp.

three reported

349-353.

namely: Multi-layer Neural Network, ADTree and J4.8 tree. The one line segment method

[7]

K. Adams, Optimisation of Multiple-Valued Logic Polynomials by Polarities and Affine Transforms, Pd.D. Dissertation, University of Ulster, Faculty of Informatics, at Magee College Londonderry, 2007.

[8]

K. Adams, D. Bell, L.P. Maguire, R. J.

tailed at the end. The heuristics demonstrated here have shown to be comparable to and often better than the more complicated ideas of neural network and decision trees. This fundamentally different

McGregor, (2003), “Knowledge discovery from decision tables by the use of multiplevalued logic,” Artificial Intelligence Review,

and fresh approach should now be explored

Vol. 19, No. 2, 2002, pp. 153-176.

further on a wider set of benchmarks.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

178


[9]

X. Deng, and Z. Wang, “ Learning probability

[18]

L. Wending, J. Rendek

and P. Masksakis,

distributions of signed fuzzy measures by

“Selection of suitable decision rules using

genetic algorithm and multipleregression”,

Choquet integral”, in SSPR&SPR 2008, LNCS

Proc. IFSA 2005, pp. 438-444.

5342,: N. D. Vitora Lobo et al. (Eds.): Springer-Verlag Berlin Heidelberg 2008, pp.

[10]

https://archive.ics.uci.edu/ml/datasets

947–955.

/Breast+Cancer+Wisconsin+ (Original). [19] [11]

O. L. Mangasarian and W. H. Wolberg:

K. Xu., Z. Wang, Y. Ke, “Calssification by nonlinear integral projections”, IEEE Trans. On

"Cancer diagnosis via linear programming",

Fuzzy Systems, 2003, Vol. 1, No. 2, pp 187-

SIAM News, Vol. 23, Number 5, September

201.

1990, pp. 1-18. [20] [12]

W. H. William, and O.L. Mangasarian:

D. B. Fogel, E. C. Wasson, E. M. Boughton, “Evolving neural networks for detecting breast

"Multisurface method of pattern separation for

cancer”, Cancer let. 1995; 96(1): pp. 49-53.

medical diagnosis applied to breast cytology", Proceedings of the National Academy of

[13]

[21]

H. A. Abbass, M. Towaey, G.D. Finn, “C-net: a

Sciences, U.S.A., Vol. 87, December 1990, pp.

method for generating non-deterministic and

9193-9196.

dynamic

multivariate

decision

trees”,

Knowledge Inf. Syst. 2001, No 3: pp. 184-97.

G. Choquet, “Theory of Capacities”, Annales de l’Institut Fourier, 1953,Vol. 5, pp. 131-295, (in French). [22]

[14]

[15]

[16]

H. A. Abbass HA, “An evolutionary artificial

Z. Yang, R. Yang and K, Leung, Nonlinear

neural networks approach for breast cancer

Integrals and their Applications in Data

diagnosis”, Artificial Intelligence in Medicine

Mining: World Scientific, 2010.

2002, No. 25, pp. 265-281.

J. Marichal, “Entropy of discrete Choquet

[23]

A. Aloraina, “Different machine learning algorithms

Vol. 9, No 1, pp. 164–172.

International Journal of Artificial Intelligence

T. Demirel

for

breast

cancer

diagnosis”,

capacities”, IEEE Trans. Fuzzy Syst, 2001,

& Applications (IJAIA), Vol. 3, No. 6,

and E. Taskan, “Multi-criteria

November 2012, pp. 21-30.

evaluation of shifts and overtime strategies using Choquet integral”, Proceedings of the

[17]

[24]

A. Roberts, Guide to Weka, [on line]. UK. Available

2012, July 4-6, London U.K.

roberts.net/teaching/ai32/weka.pdf.

J.

Hui,

and

Z.

Wang,

from:

http://www.andy-

World Congress on Engineering, vol. 2, WCE

“Nonlinear

multiregression based on Choquet integral for data with both numerical and categorical attributes”, Proc. IFSA 2005, pp. 445-449.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

179


Study on the Optimal Design of AMP Body Frame Yoon Gi-Ho and Yeom kyung-won and Kim Soo-Yong and Lee Hae-Soo and Lim kyung-Hee JUNG-A MARINE CO., LTD. JUNG-A MARINE 1469-5, SONGJEONG-DONG, BUSAN, KOREA [email protected]

ABSTRACT The amount of exhausts of Sox and Nox from a vessel would be equivalent to the amount from 50 diesel trucks, and the total amount of exhausts from LA harbor in U.S.A. is 31.4 ton/day which equals with the amount from 31 million cars. To regulate these exhausts the Act of POLA (Port Of Los Angeles) Green Port was legislated based on the NEI (No Emission Increase) policy of America, and with the initiation of Tokyo Protocol in Asia the global strategy toward Green Port is being implemented. That is, the demand for electric power supply cable to activate special vessels in port, cranes, and other appurtenant facilities is now explosively increasing with the regulation of oils for big equipment, and along with this trend the ground power supply equipment for AMP (Alternative Maritime Power) Ships is being developed. For the case of maritime vessels the amount of fuel consumption can be reduced by decreasing the weights of machinery and materials. And in this paper, the structural stability and optimal configuration of the ‘Body Frame’ of ground power supply equipment for ships will be examined through the commercial analytical program.

KEYWORDS AMP (Alternative Maritime Power), Body Frame, Thickness, Optimization Design

and developed countries are seeking to develop various reduction strategies against air pollution. However Korea has recently conducted several studies concerning air pollution in port industry. The main purpose of the paper is to suggest emission reduction strategies for bulk terminal in Port of Incheon, which handles large amount bulk cargoes as a gateway for the metropolitan area. For this aim, the clean air strategies of the world major ports were considered and air pollution reduction strategies were suggested. The main findings of this paper are as follows. First, the emission reduction strategies for container terminal are should be integrated based on technologies changes, operational changes and market-based measures. Second, the emission reduction strategies for bulk terminal can be effective when use innovative measures during loading, unloading and storage process such as telescopic cascade trimming chute, snake sandwich equipment, dry fog system and dome structure. Finally, investigation on actual conditions of air pollution in Korean ports and development of environmental evaluation scheme for persisting monitoring should be conducted. [1]-[4]

1 INTRODUCTION In the energy-climate era, pollution emissions from port activities have a significant issue in international shipping and port community. Thus international organization such as IMO

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

2 HEADINGS The ground power supply equipment for ships comprises the Cable Reel that suspends the cable, the Body Frame which supports the

180


Control Panel Cable Reel, the Guide Roller to prevent the twisting of cable and to protect the covering, the Plug to connect the anchored ship to the electrical port, and the Hand Gear enabling the vertical adjustment of the Guide Roller. The schematic diagram of components is illustrated in Figure 1. The material of the Body Frame to be employed for the structural analysis is represented in Table 1.

Fig. 2 Boundary condition of Body Frame

2-2 The Selection of Model To reduce the weight of Body Frame the structural analysis was carried out by reducing the area of the surface suspending the Reel. The four models are illustrated in Figure 3.

Fig. 1 Composition of AMP Table 1 Material Properties of STS304 STS304

Density (kgm^-3)

Young’s Modulus (MPa)

Poisson’s Ratio

Yield Strength (MPa)

Value

8000

1.93E+5

0.29

215

(a) Body Frame_1

(b) Body Frame_2

2-1 Boundary Conditions for the Analysis The lower part of Body Frame to be fixed on the ship by welding and the Side Frame holding each frame were fixed in the analysis by applying the condition of ‘Fix’, and to the part the reel would be suspended on the load of about 3 ton was applied by multiplying the value added the weights of Reel and Cable by 1.2. The value of Acceleration was not considered in this analysis since the AMP Cable Reel would be used when a ship is anchored in a harbor. The boundary conditions of the Body Frame are illustrated in Figure 2.

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

(c) Body Frame_3 (d) Body Frame_4 Fig 3. Body Frame of Four Model

The Body Frame having the safety factor over 2.0 with minimized weight which could bear the weighs of Reel and Cable among components of the ground power supply equipment for ships is to be designed.

181


(b) Stress Value of The body Frame_2

Fig .5 When the thickness of the surface stress analysis result 4mm

(c) Stress Value of (d) Stress Value of The body Frame_3 The body Frame_4 Fig 4. Stress Analysis Results by Weight

Fig. 6 Safety Factor When the thickness of the surface-4mm

(a) Stress Value of The body Frame_1

From results of the analysis on four models the model having the least weight (364kg) was selected since values of the Stress and Safety Factor did not show significant differences. 2-3 The Analytical Results with Varied Thickness of Plate To determine the thickness of plate along with the reduction of surface area of selected model the analysis was carried out by applying the value of 0.4mm spacing from 2mm to 4mm, and the resulting values of Safety Factor and Stress were identified. Figure 5-16 illustrates the analytical results of Stress and Safety Factor at 4mm of thickness of the plate.

Fig .7 When the thickness of the surface stress analysis result 3.6mm

Fig. 8 Safety Factor When the thickness of the surface-3.6mm

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

182


Fig .9 When the thickness of the surface stress analysis result 3.2mm


Fig. 11 When the thickness of the surface stress analysis result 2.8mm


ISBN: 978-1-9491968-07-9 ©2015 SDIWC

Fig. 13 When the thickness of the surface stress analysis result 2.4mm


Fig. 15When the thickness of the surface stress analysis result 2mm

Fig. 16 Safety Factor When the thickness of the surface-2mm

183


3 RESULT

4 CONCLUSIONS

Fig. 17 Frame by Weight

1. The model of total weight of 364kg which satisfies the Stress of the item with maximally reduced area of the surface was selected. 2. The thickness of Body Frame was determined to be 2.4mm. 3. The value of Stress corresponding to the thickness of 2.4mm of the Body Frame was 101.24MPa. 4. The value of Safety Factor corresponding to the thickness of 2.4mm of the Body Frame was 2.12. 5. Since the minimum Safety Factor was set as 2.0 for the design of Body Frame the optimal thickness of the plate was determined to be 2.4mm. 5 ACKNOWLEDGEMENT

Fig. 18 Frame by Stress Analysis Results

This work was supported by grants of Industry and Energy Ministry super material convergence products industrialization technology development (No. 1415129188). 6 REFERENCES [1]

C. H. Han, Journal of Korea Port economic Association, Vol.27, No.1, PP. 281-304, 2001.

[2]

AAPA. Green Ports : Environmental Management and Technology at U. S. Ports, 2001

[3]

D, Bailey. and G, Solomon, “Pollution Prevention at Ports : clearing the air,” Enviromental Impact Assessment Review, Vol.24, PP.749-774, 2004.

[4]

S. B. Bill., T. M. Guire., D. Calkins., Environmental Impacs of international shipping : Case Study of the Prots of Los Angeles and Long Beach, OECD, 2010.

[5]

M. S. Choi, D. J. Jang, D. J. Yeo, and D. H. Moon, The Korean Society for power System Engineering ., “Structure analysis of plate Structures by trantfer of Stiffness Coefficient,” PP.85-90, June, 2006.

[6]

Y.S. Kim, J.G. Park, D. C. Ahn, and Y. H. Kim, “review of Formability and Formaing Property for Stainless Steel”, Transactions of Materials Processing, , Vol.20, No.3, PP193-205, 2011.

Fig. 19Thickness by Stress Analysis Results

Fig. 20 Thickness by Safety Factor Results

ISBN: 978-1-9491968-07-9 ©2015 SDIWC

184