An Energy Efficient MAC Protocol for Cluster Based ... - IEEE Xplore

43 downloads 364079 Views 215KB Size Report
Technical Education Faculty. Sakarya University. Sakarya, Turkey [email protected]. Abstract: Energy efficiency is one of the most important design.
An Energy Efficient MAC Protocol for Cluster Based Event Driven WSN Applications Nukhet Sazak

Ismail Erturk

Etem Koklukaya

Murat Cakiroglu

Engineering Faculty Sakarya University Sakarya, Turkey [email protected]

Technical Education Faculty Kocaeli University Kocaeli, Turkey [email protected]

Engineering Faculty Sakarya University Sakarya, Turkey [email protected]

Technical Education Faculty Sakarya University Sakarya, Turkey [email protected]

Abstract: Energy efficiency is one of the most important design criteria for long lasting WSN applications. There are many interesting works related to Medium Access Control (MAC) protocols proposed for low energy consumption and high system performance in the literature. TDMA based MAC protocols are more energy-efficient than their traditional contention based counterparts. However, conventional TDMA based MAC protocols are not well suited to the event driven WSN applications. There are some MAC protocols allocating time slots to only source nodes. The distinctive feature of our proposed MAC protocol is that it assigns a time slot to only one of the source nodes all with the same data sensed and to be sent. Thus it reduces data transmission redundancy and achieves energy savings. The analytical performance metrics and expressions are obtained for the proposed MAC protocol and its performance is compared to the conventional TDMA, E-TDMA, BMA and EA-TDMA protocols.

1. INTRODUCTION A wireless sensor network (WSN) consists of numerous low-power, low-cost and multi-functional sensor nodes which can be densely deployed either inside the phenomenon or very close to it [1]. The WSN applications are so diverse that the application range is only limited by the imagination [2]. In most of these applications, a large number of sensor nodes are deployed in harsh networking environments. It is difficult and even impossible to recharge or replace the batteries of the nodes [3]. Thus energy efficiency becomes one of the most important design consideration for WSNs. Energy efficient solutions are proposed at all layers of communication stack. However, there are a lot of medium access control (MAC) protocols considering energy efficiency as the primary goal in the literature since MAC has direct control over the radio that is the most energy consuming part of a sensor node [3]. A WSN MAC protocol coordinates how and when its nodes access the wireless channel to be able to use the shared medium effectively and fairly. Idle listening, collision, overhearing and control packet overhead are the sources of energy waste [4]. Collision occurs if two or more nodes try to

transmit data at the same time. When a sensor node listens the channel to check whether there is data related to it, idle listening occurs. Overhearing takes place when a node receives messages that are destined to other nodes. Transmitting control packets also causes energy waste since these packets do not convey useful data. These issues should be carefully taken into account in designing an energyefficient MAC protocol. MAC protocols are usually divided into two classes as contention-based and schedule-based. Contention-based (random or unscheduled) protocols allow nodes accessing the medium independently from any other. They are inefficient in terms of use of energy sources because they have all of energy waste [5]. Schedule-based (collision-free) protocols are more energy-efficient than contention-based ones. The FDMA and CDMA schemes are inappropriate protocols for WSNs due to their extra circuitry and computational complexity requirements, respectively. Most of the scheduled based protocols used for WSNs are a variant of time division multiple access (TDMA) [2]. TDMA-based protocols reduce the duty cycle of sensor nodes and all of the energy waste sources mentioned above are avoided or diminished because nodes transmit or receive in their own allocated slots [3]. Clustering techniques are used against the limited scalability and adaptability disadvantages of TDMAbased protocols limiting their usage for WSNs. When traditional TDMA is implemented in a cluster consisting of a cluster head and member nodes, member nodes turn on the radio in their allocated slots and transmit the data. In conventional TDMA, slots are allocated to every node no matter they have data or not to transmit. The problem is that each node has to turn on its radio in its allocated time slot and remains idle even if it has no transmission. A sensor node has four operating modes: transmit, receive, idle and sleep. The energy consumption in an idle mode is approximately same as receive mode [5]. However, the sensor node consumes very low power in the sleep mode. Thus the more the sensor node sleeps, the more its energy is saved. Overcoming this problem, the solution is a slot assignment to only source nodes which mean they have data to send. There are some

papers aiming at time slot assignments only to source nodes (BMA etc.). When the specific conditions occur according to the application requirements (e.g. temperature threshold exceeds), nodes are prompted to transmit the related measurement values in event driven WSN applications. The neighbor nodes close to the phenomenon may sense and also might have to transmit the same data. This data redundancy consequently results in energy waste. Then the slot assignment to only source nodes is not the optimum solution. A more energy efficient protocol is achieved by allocating time slots to the only one of the nodes with same data. The aim of this presented study is to propose a new energy efficient TDMA-based MAC protocol by assigning slots to only the nodes that have different data in cluster based event driven WSN applications. In this paper, a new energy efficient TDMA-based MAC protocol is presented in which same data transmitted by more than one node is prevented in event driven WSN applications. In section 2, conventional TDMA, E-TDMA, BMA and EATDMA protocols are briefly explained. The operation of the proposed protocol is defined in section 3. In section 4, the analytical expressions are given for the proposed protocol and the other protocols mentioned in related work. The energy consumption of the protocols is compared in the graphs in section 5. Finally, the contribution of the paper is emphasized and information about the further study is given in section 6. 2. TDMA-BASED MAC PROTOCOLS FOR WSNS In conventional TDMA protocol, each round includes a set-up and a steady-state phase. After the cluster formation each node turns on its transceiver in its allocated slot sent by the cluster head (CH). If it has data, it transmits; otherwise it is in idle mode. In energy-efficient TDMA (E-TDMA) protocol, the energy consumption in idle mode is reduced by letting the nonsource nodes turn off their radios in their allocated slots [6]. Bit-map assisted (BMA) [7], [8] is also schedule-based MAC protocol. BMA is designed for event driven WSN applications in which sensor nodes transmit data only when they observe significant/predefined events. The operation of BMA is also divided into rounds. Each round includes a setup and a steady-state phase. The steady-state phase is partitioned into sessions consisting of contention period, data transmission period and idle period (Figure 1). The size of the data transmission period is variable since each node does not have always data to send. On the other hand, the sum of the data transmission and idle periods is equal to a constant value in each session. All of the nodes keeps their radios on during every contention period. The contention period follows a TDMA-like schedule. Each node is assigned a time slot and it transmits 1-bit control message if it has data to send.

Otherwise, that slot remains empty. After contention period is completed, the CH knows the source nodes, prepares a transmission schedule and broadcasts it. Then the system enters the data transmission period. If none of the member nodes has data to transmit, the system enters the idle period which lasts until the next session. Throughout data transmission period, source nodes turn on their radios and transmit the data to the CH over their allocated slots. The other member nodes keep their radios off during data transmission period. In idle period, the radios of all nodes are off. BMA is an appropriate protocol for low traffic conditions since it delivers better performance than TDMA and ETDMA for low and medium traffic loads. In Energy-Efficient Adaptive TDMA (EA-TDMA) protocol [9], CH broadcasts the schedule to all nodes after cluster formation in set-up phase. In EA-TDMA, each node wakes up in its allocated slot, checks its buffer and transmits the data if it has. If the node has no data in its own slot, it immediately turns the radio off. In the absence of data, it is in sleep mode instead of idle mode. 3. THE PROPOSED MAC PROTOCOL A content based scheduling approach has been presented for reducing the number of nodes transmitting data [10]. The proposed MAC protocol inspiring from this work for WSNs consists of rounds including set-up and steady-state phases as in the other classical scheduled protocols. A cluster head is chosen and the cluster is formed according to a specific mechanism in the set-up phase, There is also a contention period as in BMA in the steady state phase (Figure 2). However, source nodes transmit the difference data between the threshold and measurement values instead of declaring whether they have data or not by sending 1-bit in BMA. A 4bit time slot is allocated to each node. When the nodes sense the values equal or bigger than the predefined threshold value, called as source nodes, they transmit the difference data between the threshold and measurement values in those 4-bit time slots [10]. Consequently, the cluster head not only knows the source nodes but also the nodes with the same data and assigns data slots according to this knowledge. The proposed MAC application is assumed for such networking environments that include sensors closed to each other and therefore may sense almost similar values. As a result, for example, in a fire situation two different sensors’ readings would be not to much different and 4 bits are long enough to express it. On the other hand, this can be extended (e.g. to 6 bits) with the expense of a small increase in the node energy consumption. Forest fire can be given as an example for the event driven WSN applications. A large number of nodes with predetermined temperature threshold is deployed randomly and densely by a plane over the forest. A source node is defined as the node which measures at least the threshold value. During the contention period the source nodes transmit

the difference data and the non-source nodes remain in idle mode in their own 4-bit time slots. The cluster head compares the difference data from the source nodes and assigns data slot to the only one of the nodes with the same data. In this way, different temperature measurements are transmitted in data transmission period after the contention period.

transmit or receive a control packet. The number of frames in a round is expressed by l. 4.1 Basic TDMA Protocol In set-up phase, cluster head and all nodes turn on their radios. The CH determines the time slots to each node and broadcasts it to them. Energy consumed by each node to receive a control packet is PrTc and energy consumed by the CH to transmit a control packet is PtTc. Thus total energy consumption [9] in the set-up phase is calculated as in Equation (2). EC= Pt Tc + NPrTc

(2)

In steady-state phase, energy consumption of a source node Figure 1 - Illustration of single round for BMA protocol [7].

is Edn= PtTd

(3)

and energy consumption of a non-source node is Ein= PiTd

(4)

Energy consumed by the CH in this phase is Ech= nPrTd+(N-n)PiTd Figure 2 - Illustration of single round for our proposed protocol[10].

As a result total energy consumption during a round in a basic TDMA protocol is found as [9] ETDMA= PtTc+NPrTc+l[nPtTd+2(N-n)PiTd+nPrTd]

4. ANALYTICAL MODEL OF THE PROPOSED MAC PROTOCOL We suppose a clustered network with one cluster head (CH) and N member (non-CH) nodes in a cluster. Energy consumption is calculated over one round. Each round includes k sessions/frames. The nodes which have data to transmit are called as source nodes. There are ni source nodes in the ith session/frame. Bernoulli trial is used to calculate the probability which a node has data to send or not as in BMA [7][8]. The number of source nodes in a frame is given as E[ni]= Np =n, i=1,2,…k

(1)

where p is the probability. The power consumption in the receive mode, transmit mode and idle mode are represented by Pr, Pt and Pi respectively. Ein is the energy dissipation of a node in the idle listening period. Td is the time required to transmit or receive a data packet, Tc is the time required to

(5)

(6)

4.2 E-TDMA Protocol Energy consumption in E-TDMA protocol during set-up phase is the same as in TDMA protocol. Similarly, there is no change in energy consumption values in CH and source nodes during steady-state phase. The difference results from non-source nodes which do not turn on their transceivers in their allocated slots [6]. Hence energy consumption in ETDMA protocol is given as in Equation (7). EE-TDMA= PtTc+NPrTc+l[nPtTd+(N-n)PiTd+nPrTd]

(7)

4.3 BMA Protocol Tch is the time required for the CH to transmit a control packet. Energy consumed by a source node is Edn= PtTc+(N-1)PiTc+PtTd+PrTch

(8)

Ese= nPtTd+(N-n)PeTe+(N-n)PiTd+nPrTd

and energy consumed by a non-source node is

(18)

(9)

Total energy consumed during a round in EA-TDMA protocol [9] is computed as

in the BMA frame [9]. The energy consumed by the CH is [9]

EEA-TDMA= PtTc+NPrTc+ l[nPtTd+(N-n)PeTe+(N-n)PiTd+nPrTd] (19)

Ein=NPiTc+PrTch

Ech = n(PrTc+ PrTd) + (N-n) Pi Tc + PtTch

(10)

Total energy consumed during a round in the BMA protocol [9] is given as EBMA= l[n(PtTc+(N-1)PiTc+PtTd+PrTch)+ (N-n)(NPiTc+PrTch)+n(PrTc+PrTd)+ (N-n)PiTc+PtTch]

(11)

4.4 EA-TDMA Protocol The power consumption while a non-source node turning on to check its buffer to learn whether it has data or not is defined as Pe. The time required for a non-source node to check its buffer is defined as Te [9]. In the set-up phase CH consumes Ech=PtTc

(12)

(13)

Thus total energy consumption in the set-up phase is given in [9] as EC= PtTc+NPrTc

(14)

Energy consumption expressions for a source node and non-source node in a EA-TDMA frame is given as Edn=PtTd

(15)

and Ee=PeTe

(16)

respectively [9]. As the energy consumed by the CH is Ech-e=nPrTd+(N-n)PiTd

Considering p' is the probability and n is the number of source nodes, the number of source nodes which are assigned data slots (number of source nodes with different data) is defined as m=np'

(20)

Energy consumed by a source node in the contention period in a frame is then Edn= Pt(Tc+(3N/data rate))+(N-1)Pi(Tc+(3N/data rate))+ Pr (Tch+(3N/data rate)) (21) In the data transmission period, each of the m nodes which are assigned time slots consumes in a frame is

to transmit a control packet and each node consumes En=PrTc

4.4 Our Proposed MAC Protocol

(17)

system energy dissipation in a EA-TDMA frame is found as [9]

Es=PtTd

(22)

Energy consumed by a non-source node in a frame can be expressed as Ein=NPi(Tc+(3N/data rate))+ Pr(Tch +(3N/data rate))

(23)

The CH energy consumption consists of “mPrTd” "nPr(Tc+(3N/data rate))”, “(N-n)Pi(Tc+(3N/data rate))” and “PtTch+(3N/data rate)”, thus in the data transmission period Ech= nPr(Tc+(3N/data rate))+mPrTd+ (N-n)Pi(Tc+(3N/data rate))+ Pt(Tch+(3N/data rate))

(24)

Total energy consumed during a round in the proposed protocol is finally expressed as EPROPOSED= l[n{Pt(Tc+(3N/data rate))+ (N-1)Pi(Tc+(3N/data rate))+ Pr (Tch+(3N/data rate))}+mPtTd+ (N-n){NPi(Tc+(3N/data rate))+ Pr (Tch+(3N/data rate))}+ nPr(Tc+(3N/data rate))+ (N-n)Pi(Tc+(3N/data rate))+ mPrTd+Pt (Tch+(3N/data rate))] (25)

5. NUMERICAL ANALYSIS Rockwell’s WINS model [11] is used as in the BMA protocol for numerical analysis. In this model transmitting power is 462 mW, receiving power is 346 mW, idle listening power is 330 mW and data rate is 24 kbps for transceiver. A data packet size of 250 bytes, a control packet size of 18 bytes, Pi=Pe and Tch=Tc=Te are assumed. The graphs of energy consumption versus number of nodes and p' probabilities are shown in Figure 3 for number of frames (l) 10 and probability (p) 0.6.

energy consumption (Joule)

p=0.6 p'=0.25 25

E(TDMA)

20

E(E-TDMA)

15

E(BMA)

10

E(EA-TDMA)

5

E(PROPOSED)

0 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 N

energy consumption (Joule)

p=0.6 p'=0.5 30

E(TDMA)

25 20

E(E-TDMA)

15

E(BMA)

10

E(EA-TDMA)

5

E(PROPOSED)

0 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 N

energy consumption (Joule)

p=0.6

p'=0.75

30

E(TDMA)

25 20

E(E-TDMA)

15

E(BMA)

10

E(EA-TDMA)

5

E(PROPOSED)

0 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 N

energy consumption (Joule)

p=0.6 p'=1 35 30 25 20 15 10 5 0

E(TDMA) E(E-TDMA) E(BMA) E(EA-TDMA) E(PROPOSED) 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 N

Figure 3 - Energy consumption versus number of nodes. In energy computations, the energy consumed by the cluster head for comparing difference data from source nodes is ignored. Transmitting 1 kb data over 100 m consumes

approximately the same amount of energy as executing 3 million instructions [12]. Considering that communication is much more energy consuming than computation, it is clear that the proposed protocol will provide energy-efficient results even the energy consumption of computation is included. Although the source-to-cluster-head control message is only 1-bit long in BMA protocol, control packet includes this 1-bit control message plus other MAC level overhead information [13]. Since the contents of the control packet is unknown and it has already N bits, three times the number of nodes (3N) bits are added to the control packet length to allocate 4-bit slot for each node in computations. The p' denotes the probability what fraction of the source nodes have the different data and hence the percentage of the source nodes assigned data slots. Since slot assignment is realized without considering the contents of the data in TDMA, E-TDMA, BMA and EA-TDMA protocols, energy consumption values of these protocols do not vary with p'. The effect of the p' on our proposed protocol is seen in the graphs. The more p', the more source nodes assigned data slots and the more source nodes with different data. When 25% of the source nodes has different data, this protocol is more energy efficient than the others up to different number of nodes for each of them. For example, it outperforms TDMA and BMA up to N=17 and N=25 respectively for p'=0.25. On the other hand, it is seen that energy consumption of the proposed protocol becomes higher than the others by increasing number of nodes. The optimum number of clusters in each round for a 100-node network determined in [14] is 5 in terms of energy consumption. Assuming the uniform distribution of nodes, the number of nodes per cluster is 20. Considering the results in Figure 3, we may conclude that the proposed protocol provides a high efficiency in energy usage for cluster based event driven WSN applications. 6. CONCLUSION In this paper, we present a TDMA-based MAC protocol which offers data slot assignment by considering source nodes transmitting same data in event driven WSN applications. In this protocol, contention period is increased to transmit difference data in 4-bit time slots instead of 1-bit slot. On the other hand, energy saving is achieved since the number of assigned data slots is decreased. The energy efficiency of the proposed protocol is pointed out by comparing the other protocols. In further study, this protocol will be investigated in terms of latency. Also, we plan to do simulation studies related to the protocol of which analytical expressions are given in this paper.

REFERENCES [1] Ian F. Akyildiz, Weillian Su, Yogesh Sankarasubramaniam, and Erdal Cayirci: “A survey on sensor networks”, IEEE Communications Magazine, August 2002. [2] Kazem Sohraby, Daniel Minoli, and Taieb Znati: “Wireless sensor networks: technology, protocols, and applications”, John Wiley & Sons, pp. 75-229, 2007. [3] Kurtis Kredo II, Prasant Mohapatra: “Medium access control in wireless sensor networks”, Computer Networks, vol. 51, no. 4, pp. 961-994, 2007. [4] Wei Ye and John Heidemann: “Medium access control in wireless sensor networks”, USC/ISI Technical Report ISITR-580, www.isi.edu/~johnh/PAPERS/Ye03c.html, 2003. [5] Hnin Yu Shwe, Jiang Xiao-hong, and Susumu Horiguchi: “Energy saving in wireless sensor networks”, Journal of Communication and Computer, vol. 6, no.5, serial no. 54, May 2009. [6] Jing Li, Georgios Y. Lazarou: “Modeling the energy consumption of MAC schemes in wireless cluster-based sensor networks”, Proceedings of the 15th IASTED International Conference on Modeling and Simulation, pp. 313-318, Marina del Rey, California, USA, March 2004. [7] Jing Li, Georgios Y. Lazarou: “A bit-map assisted energy-efficient MAC scheme for wireless sensor networks”, Proceedings of 3rd International Symposium on Information Processing in Sensor Networks, pp. 55-60, Berkeley, California, USA, April 2004. [8] Jing Li: “A bit-map assisted energy-efficient MAC scheme for wireless sensor networks”, M.Sc. Thesis, Electrical Engineering, Missippi State University, May 2004. [9] G. M. Shafiullah, Adam Thompson, Peter. J. Wolfs, and Shawkat Ali: “Energy-efficient TDMA MAC protocol for wireless sensor networks applications”, Proceedings of International Workshop on Internet and Distributed Computing, Khulna, Bangladesh, December 2008. [10] Nukhet Sazak, Ismail Erturk, Murat Cakiroglu, Etem Koklukaya: “A content based scheduling approach for wireless sensor networks MACs”, International Symposium on Sustainable Development, ISSD 2010, Sarajevo, Bosnia and Herzegovina, June 8-9 2010. [11] Vijay Raghunathan, Curt Schurgers, Sung Park, and Mani B. Srivastava, “Energy-aware wireless microsensor networks”, IEEE Signal Processing Magazine, pp. 40-50, March 2002. [12] Gregory J. Pottie, William J. Kaiser: “Wireless integrated network sensors”, Communications of the ACM, 43(5), pp. 51-58, May 2000. [13] Georgios Y. Lazarou, Jing Li, Joseph Picone: “A clusterbased power-efficient MAC scheme for event-driven sensing applications”, Ad Hoc Networks, Elsevier, vol. 5, issue 7, pp. 1017-1030, September 2007. [14] Wendi B. Heinzelman, Anantha P. Chandrakasan, Hari Balakrishnan: “An application-specific protocol architecture

for wireless microsensor networks”, IEEE On Wireless Communications Trans. vol. 1, no. 4, pp. 660-670, October 2002.