A Fuzzy Logic Based Cross-Layer Mechanism for ... - IEEE Xplore

2015 IEEE 26th Annual International Symposium on Personal, Indoor and Mobile Radio Communications - (PIMRC): MAC and Cross-Layer Design

A Fuzzy Logic Based Cross-Layer Mechanism for Medium Access Control in WBAN Seyed Mohammad Nekooei

Gang Chen

Ramesh Kumar Rayudu

School of Engineering and Computer Science Victoria University of Wellington Wellington, New Zealand Email: [email protected]



Abstract—Over the past decade, advances in electronics, computer science, and wireless technologies have brought Wireless Body Area Network (WBAN) into many interesting applications. Particularly for the healthcare application, reliability is considered as a very important aspect for WBAN. Being the main focus of this paper, we aim at improving reliability by reducing the collision rate and increasing the packet delivery ratio. We also strive to enhance the performance in terms of throughput in the WBAN while maintaining message latency at a reasonable level. In an effort to achieve these goals, we introduce a new cross-layer fuzzy logic based backoff mechanism. Through this method, instead of relying merely on Medium Access Control (MAC), information from physical and application layers will be exploited as well. Moreover, since independent decision making is supported by each sensor without relying on any coordinating devices, communication in the WBAN becomes very flexible. Specifically, rather than determining Backoff Exponent (BE) in IEEE 802.15.4 through a blind try-and-error process, the proposed fuzzy logic system determines the BE by considering both the current channel condition and application requirements. This feature gives the proposed system a higher level of adaptability. The simulation results clearly show noticeable improvement in reliability and performance without significantly increasing the message latency. Keywords—Cross layer, Fuzzy Logic, MAC layer, CSMA/CA, Wireless Body Area Network (WBAN)

I.

I NTRODUCTION

Wireless Body Area Network (WBAN) is a recent leap in information technology [1]. It is comprised of multiple devices (also known as network nodes) placed in different locations on the human body and is frequently utilised to remove the wires connecting these nodes. It brings more freedom to realtime monitoring, especially for ambulatory patients in hospital, home, or mobile health monitoring situations. Each WBAN has one and only one special node known as the coordinator [2]. The most common standard used in WBANs is IEEE 802.15.4 [1], [2], which is developed to support applications with low power and low cost requirements. IEEE 802.15.4 has been evaluated recently and it has been shown that the reliability of this IEEE standard in terms of Packet Delivery Ratio (PDR) and collision rate can be very limited [3]. As a result, a WBAN may not be suitable for applications with high sensor densities [3]. In practice, reliability is an important requirement. For

978-1-4673-6782-0/15/$31.00 ©2015 IEEE

1094

instance, in healthcare monitoring applications, reliable communication is essential in order to bring dependable medical care to patients. Since WBANs are used to carry medical data and signal emergencies like vital organ failures, in worstcase scenarios, communication failure may cause death when a life-threatening event has been left unattended. Driven by this understanding, some exciting research has already been conducted to enhance the level of PDR [3], [4], [5]. Although achieving 100% reliability is ideal, this is not possible in practice. Because of that, the aim of this paper is to improve the reliability of the current standard. In addition to reliability, performance in terms of throughput poses another issue for IEEE 802.15.4 [3]. Particularly for the healthcare monitoring applications, WBANs are required to achieve high throughput in order to support sensors with high transmission rate such ECG sensors. For such applications to be successful, Chen et al. [5] proposed a soft-computing technique to improve both network reliability and performance. A similar work was also conducted by Mouzehkesh et al. [4] by proposing Dynamic delayed Medium Access Control (D2 MAC) to improve the network reliability and performance of IEEE 802.15.4. However, reliability and performance improvement should not be realised at the cost of significantly increasing the message latency. As a matter of fact, the average message latency must be kept at a reasonably low level. This requirement can be easily understood in medical applications where timely communication is highly desirable. In order to decrease the network delay, Additional Carrier Sensing (ACS) method [6] has been proposed to use the third Channel Clear Assessment (CCA) to reduce unnecessary delay introduced by the shifted backoff period. To address the technical challenges highlighted above, we will focus primarily on re-engineering the standard slotted Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) in IEEE 802.15.4. Further performance and reliability improvement can potentially be achieved by combining CSMA/CA and Guaranteed Time Slots (GTS). However, we focus preliminarily on CSMA/CA in this paper. The potential use of GTS will be considered in the future research. Our research shows that, by improving CSMA/CA alone, higher levels of network reliability and performance can be achieved. This technological improvement is complementary with use of


GTS. The current description of CSMA/CA relies on the use of a control parameter called the Backoff Exponent (BE). The maximum backoff delay is subsequently determined as 2BE . Since BE will be increased linearly whenever a collision is detected, IEEE 802.15.4 forces communicating nodes in a WBAN to increase their backoff delays exponentially without considering the application-specific requirements (ref. III-A). Different from this standard approach, in this paper, BE will be dynamically adjusted through a new cross layer approach. The adjustment is undertaken by a fuzzy logic system designed by us. We have also conducted a manual process to fine-tune fuzzy rules for the purpose of making it more suitable for healthcare applications. This research makes three contributions. (1) We successfully extend the standard to facilitate fuzzy logic based MAC and make it more competent for various WBAN-based applications. (2) Without changing the fundamental structure of the IEEE 802.15.4, the fuzzy logic preserves backward compatibility with standard CSMA/CA. (3) The final contribution is to demonstrate clearly the potential usefulness of the new scheme with comparison to several recently proposed schemes through the simulations studies. The remainder of this paper is organised as follows. Section II provides a quick overview of the default IEEE 802.15.4 backoff strategy. Section III introduces the cross-layer fuzzy logic based backoff algorithm. The fuzzy rule tuning process is presented in section IV. Section V describes the simulation implementation and results in detail. Finally, we conclude this paper and highlight the future work in section VI. II.

IEEE 802.15.4 MAC OVERVIEW

To build the technical foundation of our research, this section gives a brief overview of the current CSMA/CA mechanism in IEEE 802.15.4. According to [2], each node in a WBAN is able to transmit data to the coordinator by using CSMA/CA during the Contention Access Period (CAP). The whole procedure is depicted in Fig. 1. When the condition of fuzzyEnable in Fig. 1 is true, the flow chart shows the procedure of our proposed algorithm. Otherwise, it illustrates the standard slotted CSMA/CA in IEEE 802.15.4. In the slotted CSMA/CA mode, each sensor node holds three variables, which are the Number of Backoffs (NB), the length of contention window (CW), and the Backoff Exponent (BE). Before any new transmission, NB, CW and, BE are initialised to zero, two and, macM inBE (=3) respectively [2]. In order to avoid collision, a random waiting delay is generated uniformly at random within the range [0, 2BE − 1] before assessing the channel. When the delay elapses, CCA assesses whether the channel is busy or idle. If the channel is idle, CW is decremented by one and another CCA is performed until CW reaches zero. On the other hand, in the event of a busy channel, NB and BE are incremented by one and CW is set to two again. BE and NB cannot exceed macM axBE (default value of 5 from a valid range of [3..8]) and macM axBackof f s (default value of 4) respectively. When NB reaches the macM axBackof f s, the transmission fails. Otherwise, the backoff procedure is repeated.

1095

Fig. 1: The flow chart of CSMA/CA and CLFB algorithms where differences between the standard CSMA/CA and CLFB are highlighted

III.

C ROSS -L AYER BACKOFF M ECHANISM USING F UZZY L OGIC

As shown in Fig. 1, the adjustment of NB and BE in the IEEE 802.15.4 standard is not flexible enough to consider the varying channel condition and application-specific requirements. As a result, a node in a WBAN cannot quickly adapt its backoff delay so as to improve channel utilisation and reduce chances of collisions. To address this problem, a fuzzy logic system is developed in this research to adapt BE based on the NB-rate and the application data rate. The reason for using the two parameters will be further presented in subsections III-A and III-B. As demonstrated in Fig. 1, our fuzzy logic system promotes a cross-layer approach for adaptive media access control. Particularly, whenever fuzzyEnable is true, the fuzzy backoff bound, which is called fuzzyBE in this paper, will be obtained by using our cross layer fuzzy logic based backoff system (CLFB). CLFB requires both NB-rate and application data rate to determine the fuzzyBE. Accordingly, the random delay before assessing the channel will be within the range of [0, 2f uzzyBE − 1] where the fuzzyBE ranges from 2 to 7. In this research, we will use a popular fuzzy-inference engine known as the Mamdani fuzzy system [7] to build our CLFB. Meanwhile, the output from the fuzzy engine will be further defuzzified in order to generate a crisp fuzzyBE. Following a common practice, we use the centre of gravity as the defuzzification method. A. Fuzzy Input and Output Variables The first input variable of the CLFB is the NB-rate, which measures the average NB level over time. We believe that NB


in the CSMA/CA mechanism is a suitable indication of the channel condition because it shows whether the channel has been busy or idle in the recent past. In particular, NB will increase whenever the channel is in use. When it exceeds macM axBackof f s, the corresponding transmission effort will fail. In order to use the history of the channel to predict the imminent channel condition, we define NB-rate as the moving average of NB [8], which is given below N BRate (t + 1) = δ × N B + (1 − δ) × N BRate (t)

(1)

where δ is the discount factor. The NB-rate in our fuzzy system is categorised into four fuzzy levels, i.e. low, medium, medium-high and high. We choose to use four different fuzzy levels because a lot of related research [3], [4], [5] prefers to use at most four fuzzy levels for any types of input to a fuzzy system. However, this is not a technical restriction. In the future, we will study the possible benefits of using more fine-grained fuzzy levels. In the calculation of NB-rate (Eq. 1), a punishment will also be applied. Specifically, when a transmission attempt fails because NB exceeds macM axBackof f s, our algorithm introduces the punishment to clearly distinguish between a normal failure during CCAs and a complete transmission failure. The punishment is represented by the NB value of 6, which is clearly higher than the maximum possible NB value during any CCAs. Due to this punishment method, the full possible range of NB-rate is from zero to six. The data-rate is the second input variable to our CLFB. It is an important variable since it helps to balance between the waiting time and the channel condition. To cope with various network arrangements, we decided to normalise the data-rate. Upon joining the WBAN, a sensor node will send its application data-rate to the coordinator, which will then determine the maximum data-rate among all individual sensors in the network. The coordinator will subsequently inform the whole network about the maximum data-rate. Each sensor node will normalise its own data-rate within the range of 1 to 100 accordingly. One potential benefit of using normalised data-rate is to improve the utilisation of the network capacity. Similar with the NB-rate, the data-rate input in our CLFB is also associated with four different fuzzy levels: low, medium, medium-high and high. In addition to the inputs, the fuzzyBE, which is the output of our CLFB, is categorised by four different levels of delays, i.e. {B1, B2, B3, B4}. The complete range for BE, which is covered jointly by all the four different fuzzy levels, is from two to seven. The membership function for all input and output variables are triangular [3], [4], [5]. Fig. 2 presents an example triangular membership function used in our CLFB. In subsection III-B, we will explain how the membership functions will be manually tuned in this research. Obviously we prefer to tune a triangle function because it is simple. B. Fuzzy Logic Rules

Fig. 2: An example of triangular membership functions and tuning by adjusting the membership function parameters

fuzzyBE, will also be specified. A decision table is a handy way to present a group of rules, as shown in Table I. TABLE I: Fuzzy Logic rules for CLFB Rule No. 1 2 3 4 5 6 7 8 9 10 10 12 13 14 15 16

Condition or NB Rate Low Low Low Low Medium Medium Medium Medium Medium High Medium High Medium High Medium High High High High High

1096

Action or Consequent fuzzyBE B1 B1 B1 B2 B2 B2 B2 B3 B3 B3 B3 B4 B3 B4 B4 B4

According to Table I, moderate values of fuzzyBE are recommended when NB-rate is low. In comparison, when the channel has a busy history (i.e. NB-rate is high), greater fuzzyBE is allocated. Specifically, because the nodes with higher data rate need to access the channel more frequently, they will receive higher fuzzyBE to have a higher delay. In this way, they have a better chance of avoiding current channel congestion. As an example, in the condition of medium high NB-rate, which indicates a congested condition, rule R(12) produces higher fuzzyBE than R(9) , R(10) and R(11) . However, when the network is highly congested (i.e. NBrate is high), all nodes except low data rate nodes should receive the highest fuzzyBE to effectively decrease the chance of collision. Low data rate nodes receive less fuzzyBE to prevent being blocked by higher data rate nodes. Therefore, rule R(13) receives smaller fuzzyBE than rules R(14) , R(15) and R(16) . IV.

Since NB-rate and data-rate each have four different fuzzy levels, there are 4×4 = 16 different rule antecedents. A total of 16 separate fuzzy rules are subsequently set up in our CLFB. For each rule, a fuzzy level for the output variable, i.e. the

Antecedent Data Rate Low Medium Medium High High Low Medium Medium High High Low Medium Medium High High Low Medium Medium High High

F UZZY RULE T UNING P ROCESS

In practice, the effectiveness of CLFB depends heavily on its control rules and membership functions. Therefore, the tuning and adjustment of membership functions become an important part of CLFB design. At this stage of the research,


because the total rule set is reasonably small, we tune and adjust CLFB manually as explained below. One of the most common ways of tuning membership functions is to adjust the basic parameters defining them. For this purpose, two tuning parameters, i.e. α and β as shown in Fig. 2, are considered. In fact, α controls the shortening and widening of the membership function and β moves the membership functions to the left and right. The aim of tuning is to adjust the two parameters in a direction of improving WBAN reliability and performance.

(a) NB-Rate

(b) Data-Rate

In this research, we have three steps in the tuning process. They are detailed below, 1)

2)

3)

Tuning the output membership functions: because the output membership functions have more influence on the performance of CLFB, the adjustment is carried out on them first. Particularly the input membership functions are held steady while tuning the output membership functions. Upon adjusting a membership function, we will first adjust parameter β. After each adjustment 30 independent evaluations will be performed to confirm its usefulness, i.e. improving both reliability and performance of IEEE 802.15.4. After three local adjustments, we will then proceed to adjust α by following an identical process. After changing α and β individually, we have a good approximation of α and β. Then we will use them to perform several steps of random local search to finally determine the values for α and β. This process will be applied on all B1, B2, B3, and B4 membership functions. Tuning the input membership functions: In order to investigate suitable settings for input membership function and further improve performance of CLFB, the input membership functions are tuned and the output membership functions remain unchanged. For each membership function, tuning of α and β follows basically the procedure described in step 1. Tuning both the input and output membership functions: the previous two steps follow a greedy search strategy to find out the suitable settings for each membership function. In this final step, in order to improve CLFB performance further, a local search is performed to tune both inputs and output membership functions simultaneously.

By following this process, we have finally determined the suitable membership functions for each fuzzy level of every input and output variable. These functions are presented in Fig. 3 to make the simulation results reproducible. V.

S IMULATION I MPLEMENTATION AND R ESULTS

In order to evaluate the reliability and performance of CLFB and compare it with some existing technologies, we have performed some simulation study in this paper. In this section, detailed information and discussion related to the simulation environment, including important simulation settings, will be described in subsection V-A. The simulation results and discussions will be presented in subsection V-B.

1097

(c) fuzzyBE

Fig. 3: The tuned membership functions

A. Simulation Environment In this research, we chose to use OMNeT++ 4.4 as our simulation tool [10]. OMNeT++ is an open source network simulator framework with support for a wide variety of communication protocols. Our simulation is set up in a star-based WBAN with a single WBAN coordinator. To evaluate network conditions in different traffic loads, the number of sensors will increase from two to nine. The simulation scenarios studied here are heterogeneous since we use up to six different types of medical sensors with different data rates. A summary of different sensors and their communication features can be found in Table II (i.e. 3-lead ECG, respiratory rate, blood pressure, heart rate, motion and temperature sensors). Although these sensors only represent a small portion of all available medical sensors, they are the most frequently used sensors in many simulation studies [3]. To produce reliable results, each simulation scenario is evaluated independently 30 times. The averages of these runs will be presented as the simulation results. All body sensors are randomly placed in a 2×2 m2 region with a WBAN coordinator in the centre. Our simulation uses the IEEE 802.15.4 standard upper frequency band at 2.4 GHz, which is an important and commonly available frequency band in healthcare, with the standardised data rate of 250 Kb/s and the maximum payload size of 102 bytes. The scenarios, which are considered by our simulation, depend on the frequency band of 2.4 GHz, which is suitable for body surface to body surface communication under two possible conditions, namely Line-of-Sight (LoS) propagation and Non-Line-ofSight (NLoS) propagation. In order to create a simulation environment that closely approximates real communication situations, the log-normal shadowing model [11] is used by us to build the channel model. Existing research suggests that, in WBANs, the lognormal shadowing model can better capture the small-scale fading than the traditional Rayleigh and Ricean distributions. In this paper, “δ” in (1) is set to 0.85. Our setting comes out


TABLE II: Medical Body Sensor Specification [9], [3] Body Sensor Traffic Generation Distribution Data Rate

ECG Constant 156.25 Bps

Blood Pressure Constant 12.5 Bps

Respiratory Rate Constant 12.5 Bps

Motion Sensor Poisson 2.50 Bps

Heart Rate Constant 20 Bps

Temperature Poisson 12.5 Bps

of a trial and error process. We found that, by using it, NBrate always presents a good indication of channel condition in our simulated WBAN. In the future, more research should be performed to confirm this conjecture in a more general context.

2) MAC Throughput: Table V presents the performance of the network in terms of the MAC throughput of the network. The t-test shows that CLFB statistically outperforms CSMA/CA and ACS when there are more than two sensor nodes in the network.

B. Simulation Results

Meanwhile, when the network has nine nodes, statistically CLFB performs the same as D2 MAC and NB-step. Nevertheless, the simulation results show the throughputs of D2 MAC and NB-step in other cases outperform CLFB. That happens because they can avoid more collisions by significantly increasing the backoff delay. However, long backoff delays may not be desirable for delay sensitive applications.

In this subsection, the results obtained from the simulation are presented. The metrics, which are utilised to assess the reliability and performance of the network, are Packet Delivery Ratio (PDR) [3], Packet Latency, MAC Throughput and Number of Collisions. 1) Packet Delivery Ratio and Collision Rate: In this paper, the reliability of WBAN upon using CLFB is examined based on PDR and Number of Collisions. Table III compares PDR achieved by CLFB with several competing algorithms, i.e. CSMA/CA, ACS [6], and D2 MAC [4]. To facilitate our discussion, another competing algorithm proposed in [5] will be called “NB-Step” in the sequel. As it is shown in Table III, when there are two and three nodes in the network, PDR is nearly 100%. While the number of nodes becomes larger, PDR starts to deviate. For example, when there are nine nodes, PDR for the five algorithms will be 0.63 (CSMA/CA), 0.63 (ACS), 0.82 (D2 MAC), 0.81 (NB-step) and 0.77 (CLFB), respectively. With the help of the t-test, CLFB is shown to be statistically more reliable (i.e. higher PDR) than CSMA/CA and ACS. Meanwhile, D2 MAC and NB-step can achieve better reliability than CLFB. On the other extreme when two nodes are existing in the network, PDR is close to 100% and there are no statistically significant differences among the different algorithms. Although, CLFB can prominently improve the communication reliability in comparison to CSMA/CA nd ACS, the results show D2 MAC and NB-step can manage to achieve higher PDR than CLFB. This happens because they impose significantly longer backoffs on CSMA /CA. As a result the packet latency increases substantially. Different from the two competing algorithms, CLFB can keep the latency close to the level obtained in CSMA/CA. In the meantime, it can still achieve higher PDR (see V-B3). After investigating PDR, the number of collisions is also considered. Generally, by increasing the number of nodes, the PDR will decrease in all algorithms. This is because of the increasing number of collisions on the network, as shown in Table IV. Statistically, t-test shows that CLFB can perform significantly better than CSMA/CA and ACS when more than two nodes are involved in the simulation. On the other hand when only two nodes are simulated, no collisions can be observed. Although the results show that D2 MAC and NBstep achieved less collisions, they scarify the latency hugely. We found that the real advantage of CLFB over D2 MAC and NB-step is due to its capability of decreasing collisions without noticeably increasing communication latency from CSMA/CA.

1098

3) Packet Latency: As we mentioned earlier, the simulation results illustrated in Table VI show that D2 MAC and NBstep introduced significantly higher delays to the network than other algorithms. For example, our t-test analysis indicates that CLFB can manage to achieve significantly less delay than D2 MAC and NB-step. In fact, the delay achieved by CLFB is statistically indistinguishable from that of CSMA/CA and ACS. As we further highlighted in Table VI, in some simulations, CLFB can even achieve slightly less latency than CSMA/CA. Based on the observations above, we can claim that the main strength of CLFB lies in the fact that it can statistically dominate the performance of CSMA/CA in terms of both the reliability (e.g. PDR) and performance (e.g. latency). Particularly, among all the competing algorithms, CLFB is the only algorithm that can dominate CSMA/CA. This leads us to believe that CLFB can achieve the best utilisation of the limited communication bandwidth so as to realize a good balance among all performance metrics. VI.

C ONCLUSION

In this paper, we proposed the Cross-Layer Fuzzy logic based Backoff system (CLFB) to improve network reliability, i.e. Packet Delivery Ratio (PDR) and collision rate. We also aimed to improve the throughput in WBANs without increasing packet latency to a great extent. CLFB was designed by us to produce the Backoff Exponent (BE) by considering both the channel condition and the application data rate. This design brings higher levels of adaptability to WBANs. In addition, we also presented a manual approach to fine-tune the fuzzy membership functions in CLFB in order to enhance its effectiveness. By integrating our CLFB into the IEEE 802.15.4 MAC sub-layer, we successfully enhanced the competence of this IEEE standard for various WBAN-based applications. Moreover, this integration does not significantly change the underlying structure of the IEEE 802.15.4. Thus, backward compatibility is ensured. The results clearly showed that our CLFB achieved noticeable improvement in network reliability and performance. In the meantime, the message latency was still maintained at a reasonably low level, consistent with the original standard. One possible limitation of current research


TABLE III: Packet Delivery Ratio (PDR) from the average of 30 independent runs, the improvement of CLFB over IEEE 802.15.4 is highlighted Number of Nodes CSMA/CA (IEEE 802.15.4) ACS D2 MAC NB-Step CLFB

2 0.99 0.99 0.99 0.99 0.99

3 0.9997 0.9993 0.9997 0.9999 0.9997

4 0.81 0.81 0.97 0.98 0.87

5 0.80 0.83 0.97 0.98 0.88

6 0.80 0.79 0.95 0.94 0.85

7 0.79 0.76 0.92 0.92 0.82

8 0.70 0.70 0.88 0.87 0.80

9 0.63 0.63 0.82 0.81 0.77

TABLE IV: Number of Collisions in 600s simulation time from the average of 30 independent runs, the improvement of CLFB over IEEE 802.15.4 is highlighted Number of Nodes CSMA/CA (IEEE 802.15.4) ACS D2 MAC NB-Step CLFB

2 0.0 0.0 0.0 0.0 0.0

3 158.14 162.1 38.04 77.27 138.47

4 2127.27 2109.9 1310.94 957.9 1894.97

5 2563.7 2577.17 1563 1307.47 2131.7

6 3080.5 3259.20 2136.14 2386.07 2653.84

7 3435.14 3824.9 2211.13 2513.94 3266.27

8 4656.31 4652.73 3454.48 3272.38 3713.38

9 5644.83 5420.2 4563.5 4026.44 4798.54

TABLE V: MAC Throughput in bit per seconds from the average of 30 independent runs, the improvement of CLFB over IEEE 802.15.4 is highlighted Number of Nodes CSMA/CA (IEEE 802.15.4) ACS D2 MAC NB-Step CLFB

2 125.16 124.98 123.55 125.68 122.47

3 325.84 328.13 325.28 326.94 327.33

4 1344.58 1353.94 1700.20 1682.14 1433.80

5 1557.22 1556.53 1892.89 1867.43 1683.27

6 1698.30 1672.36 2073.39 2031.49 1853.50

7 1834.57 1690.13 2182.29 2132.80 1983.013

8 1938.53 1880.38 2460.88 2365.15 2228.30

9 2199.68 2271.13 2940.92 2845.05 2697.43

TABLE VI: Packet Latency in seconds from the average of 30 independent runs, the improvement of CLFB over other algorithms is highlighted Number of Nodes CSMA/CA (IEEE 802.15.4) ACS D2 MAC NB-Step CLFB

2 0.766 0.783 0.881 0.796 0.769

3 0.782 0.783 0.924 0.799 0.780

4 0.783 0.765 0.924 0.890 0.770

is to rely on manual tuning for CLFB. Hence, in the future, we will consider using automatic tuning based on soft computing techniques such as Particle Swarm Optimisation (PSO), Deferential Evolution (DE) and reinforcement learning algorithms, to effectively find suitable membership function in CLFB.

5 0.785 0.747 0.957 2.705 0.770

[6]

[7] [8]

R EFERENCES [1]

M. Chen, S. Gonzalez, A. Vasilakos, H. Cao, and V. C. Leung, “Body area networks: A survey,” Mobile Networks and Applications, vol. 16, no. 2, pp. 171–193, 2011. [2] “IEEE Standard for local and metropolitan area networks–part 15.4: Low-Rate Wireless Personal Area Networks (LR-WPANs),” IEEE Std 802.15.4-2011 (Revision of IEEE Std 802.15.4-2006), pp. 1–314, Sept 2011. [3] B. Otal, L. Alonso, and C. Verikoukis, “Highly reliable energy-saving mac for wireless body sensor networks in healthcare systems,” IEEE Journal on Selected Areas in Communications, vol. 27, no. 4, pp. 553– 565, 2009. [4] N. Mouzehkesh, T. Zia, S. Shafigh, and L. Zheng, “D2 MAC: Dynamic delayed medium access control (MAC) protocol with fuzzy technique for wireless body area networks,” in IEEE International Conference on Body Sensor Networks (BSN), pp. 1–6, May 2013. [5] M. Chen, X. Zhu, and H. Zhu, “Service adaptively medium access control algorithm based on fuzzy logical for energy harvesting wireless sensor networks,” Journal of Networks, vol. 9, no. 9, 2014.

1099

[9]

[10] [11]

6 0.796 0.807 1.237 2.666 0.795

7 0.844 0.814 1.341 5.990 0.892

8 0.847 1.026 2.435 7.586 0.862

9 1.073 0.985 5.821 10.538 0.928

C.-M. Wong and B.-H. Lee, “An improvement of slotted CSMA/CA algorithm in IEEE 802.15.4 medium access layer,” Wirel. Pers. Commun., vol. 63, pp. 807–822, Apr. 2012. J. M. Mendel, Uncertain Rule-Based Fuzzy Logic Systems: Introduction and New Directions. Prentice Hall, 1st ed., 2001. G. Zhou, Q. Li, J. Li, Y. Wu, S. Lin, J. Lu, C.-Y. Wan, M. D. Yarvis, and J. A. Stankovic, “Adaptive and radio-agnostic QoS for body sensor networks,” ACM Transactions on Embedded Computing Systems (TECS), vol. 10, pp. 48:1–48:34, Nov. 2011. “Health informatics - PoC medical device communication - part 00101: Guide–guidelines for the use of RF wireless technology,” IEEE Std 11073-00101-2008, pp. 1–99, 2008. “OMNeT++ network simulation framework,” December 2014. A. Taparugssanagorn, A. Rabbachin, M. Hämäläinen, J. Saloranta, J. Iinatti, and S. Member, “A review of channel modelling for wireless body area network in wireless medical communications,” in The 11th International Symposium on Wireless Personal Multimedia Communications, WPMC ’8, Sep 8-11 2008.