The Scientific World Journal

Volume 2015, Article ID 680681, 13 pages

http://dx.doi.org/10.1155/2015/680681

## Enhancing the Selection of Backoff Interval Using Fuzzy Logic over Wireless Ad Hoc Networks

^{1}Department of Information Technology, Easwari Engineering College, Chennai 600089, India^{2}Department of Computer Science and Engineering, Easwari Engineering College, Chennai 600089, India

Received 2 August 2014; Revised 21 November 2014; Accepted 6 December 2014

Academic Editor: Albert Victoire

Copyright © 2015 Radha Ranganathan and Kathiravan Kannan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

IEEE 802.11 is the de facto standard for medium access over wireless ad hoc network. The collision avoidance mechanism (i.e., random binary exponential backoff—BEB) of IEEE 802.11 DCF (distributed coordination function) is inefficient and unfair especially under heavy load. In the literature, many algorithms have been proposed to tune the contention window (CW) size. However, these algorithms make every node select its backoff interval between [0, CW] in a random and uniform manner. This randomness is incorporated to avoid collisions among the nodes. But this random backoff interval can change the optimal order and frequency of channel access among competing nodes which results in unfairness and increased delay. In this paper, we propose an algorithm that schedules the medium access in a fair and effective manner. This algorithm enhances IEEE 802.11 DCF with additional level of contention resolution that prioritizes the contending nodes according to its queue length and waiting time. Each node computes its unique backoff interval using fuzzy logic based on the input parameters collected from contending nodes through overhearing. We evaluate our algorithm against IEEE 802.11, GDCF (gentle distributed coordination function) protocols using ns-2.35 simulator and show that our algorithm achieves good performance.

#### 1. Introduction

Ad hoc network is a collection of dynamic, self-configured, and radio equipped nodes without any infrastructure. Ad hoc networks require every intermediate node to act as routers, receiving and forwarding data to every other node. This type of network is prevalently deployed in various scenarios wherein instantaneous connectivity becomes the need of the hour, either in emergency situations like a disastrous evacuation situation or in a casual get-together for presentations.

IEEE 802.11 MAC is the predominant protocol used over ad hoc networks for medium access. Binary exponential backoff algorithm (BEB) has been used by IEEE 802.11 DCF for collision avoidance. Whenever a node wants to transmit a packet, it starts sensing the medium. If the medium is idle for distributed interframe space (DIFS) period, then the node generates a backoff counter which is set for a random value between . After that, the backoff counter is decremented by one for every idle slot. If the channel is busy, the backoff counter is paused until the next DIFS free period. When the backoff counter reaches zero, the node starts transmission. Here the minimum and maximum values of CW are called and with the default values of 31 and 1023, respectively. CW is initially set to and after every unsuccessful transmission CW is doubled with the maximum limit of . Upon a successful transmission, CW is reset to .

Bianchi [1] analyzed the saturated throughput using Markov chain model and showed that the throughput increases with the smaller number of active nodes and small CW. When the number of active nodes increases smaller CW can lead to high collision. Larger CW improves the fairness among flows but reduces the overall throughput. Random BEB algorithm of IEEE 802.11 failed to improve the fairness and throughput over heavily congested ad hoc networks. Many algorithms have been developed to tune the contention window (CW) according to the congestion status. These proposals targeted for improving either throughput or fairness or both.

The algorithms proposed in the literature can be put under two categories. They are overhearing-based and non-overhearing-based solutions. In overhearing-based solutions, each node collects information such as rate, channel utilization, and buffer status from neighbors and adapts to the new contention window (CW) according to some policy. But it is well known that contention loss occurs mainly due to hidden terminals, whereas overhearing is limited to neighbors. These methods fail to consider the status of hidden terminals. Nonoverhearing solutions enable the nodes to utilize their local information like number of idle slots, busy slots, Tx failures, and so forth, to tune its contention window (CW). However, nodes within the same transmission range assess the channel in the same way. It does not help the nodes to have differentiation and get fair channel access.

The proposed solutions realize the channel congestion status either through overhearing or local information and tunes contention window size according to some policy. After tuning, they tend to select the backoff interval randomly between . In one way, this random selection helps to avoid collision between the nodes that are using the same CW. But, in another way, this randomness badly changes the order and frequency of medium access among nodes due to the zero lower bound. Random selection even leads to collisions under heavy load. This scenario affects the throughput and fairness by increasing the delay and collisions. Instead of random selection, node differentiation based on their individual parameters can optimally order the medium access among competing nodes. Optimal scheduling of the medium access is a challenging task over ad hoc networks due to its distributive and dynamic nature. Every node needs to collect information about the status of contending nodes to schedule itself to access the medium. The collected information is dynamic and vague due to the ever-changing topology of the network.

In this paper IEEE 802.11 binary exponential backoff algorithm is used for selecting contention window size (CW). To the best of our knowledge, this is the first time we introduce an algorithm that computes the backoff interval within the CW limit such that it schedules the medium in a fair and efficient way without involving much overhead. It also controls the contention by ordering the nodes according to its waiting time. We make use of fuzzy logic to compute the unique backoff interval between . Fuzzy logic is a simple problem solving methodology that accepts vague or ambiguous input values and makes us arrive at a definite and crisp output using simple if…then…rules. These rules should reflect the exact behavior of the system.

Each node collects the input parameters queue length and waiting time from contending nodes through overhearing and stores them in a neighbor table. Every node is also responsible for advertising its input parameters (queue length—myqlen and waiting time—mywt) through request to send (RTS) message. When a node overhears this RTS message, it gains knowledge about its neighbors. This learning starts from training phase and continues. During transmission each node has to choose a count between and set it for the backoff counter by comparing its own input parameters with the collected information. This dynamic and vague information is applied to the membership functions to derive fuzzy variables. These fuzzy variables are fed to the fuzzy inference engine to get fuzzy output. Finally, defuzzification helps to derive at the crisp and unique backoff counter value between .

The rest of the paper is organized as follows. In Section 2 we describe the related work and in Section 3, we brief about fuzzy logic and the steps involved and the key elements of our design in detail. We evaluate our performance against IEEE 802.11 and GDCF in Section 4 and finally we conclude our paper in Section 5.

#### 2. Related Work

BEB algorithm of IEEE 802.11 [2] suffers from severe performance degradation under heavy traffic over wireless ad hoc network. It is well accepted that contention window plays vital role in improving the aggregate throughput and fairness. In this section, we review the proposals that tune the backoff interval with the goal of achieving good throughput or fairness or both. In [3], authors derived an analytical model to find optimal value that reaches the theoretical throughput limit for* P*-persistent IEEE 802.11 protocol. To perform this, each node must have known the exact number of stations in the network and it depends on feedback information. To overcome this drawback, asymptotically optimal backoff (AOB) has been proposed by authors [4] to dynamically tune CW size according to the channel contention level. They probabilistically postpone transmission based on slot utilization factor. They show that their algorithm achieves theoretical capacity. In [5], authors tune the contention window based on the bit error rate of the medium. Both of the above methods need to estimate the number of active stations.

The authors of [6] use linear programming algorithm to optimize the minimum contention window size based on the channel condition (signal to noise ratio) and number of competing stations. Authors choose the access mode and with analytical approach to optimize the throughput. They depend on network feedback to collect the channel condition status. Virtual backoff algorithm (VBA) [7] was developed using sequencing technique to reduce the number of collisions, thereby improving the throughput. However, VBA works well only in steady state where the number of nodes is fixed. VBA suffers from collisions in a dynamic scenario. In [8], authors analytically derive contention window size based on slot utilization and optimize the throughput in both saturated and nonsaturated conditions. It utilizes only local information like busy slots and free slots and does not require estimating the number of active stations. The authors of [9] propose an algorithm GDCF wherein they perform gentle decrease of contention window to reduce collision probability. They do not reset the contention window size after every successful transmission. Instead, they find optimal counter* c* and the contention window is halved after* c* consecutive successful transmission. This method reduces the collision when the number of nodes is large. Nodes need to know the number of nodes in the network to find optimal value of* c*. Authors improve both fairness and throughput using this algorithm.

In [10], authors propose a control theoretic approach to tune contention window based on the locally available information. By comparing the average number of consecutive idle slots between two transmissions against the optimal set point, this method tunes CW and achieves optimal throughput and fairness. This method uses local information, but it also depends on the number of active stations. In [11] authors achieve fairness and weighted fairness among nodes using proposed increase with synchronized multiplicative decrease that supports background transmission. In [12], channel capacity is distributed among contending nodes through overhearing. This helps in improving the fairness among nodes. MadMAC protocol in [13] achieves both fairness and throughput using limited local information like number of experienced collisions and carrier sensing information. In [14], authors use fuzzy logic to tune the contention window based on the fuzzy parameters such as busy degree of the medium and number of neighbor nodes. This approach reduces collision probability and improves throughput and also fairness. Simplified backoff algorithm (SBA) [15] uses only local information like success, collision probability to tune CW. There are only two possible sizes for CW called (31) and (1023). CW is assigned to and during light load and heavy load, respectively. Authors claim to improve fairness and throughput. But this algorithm increases the delay due to the large CW. All of the above algorithms concentrate on tuning contention window according to the congestion level of the medium and they finally select the backoff interval randomly between .

In [16], authors change the lower bound and upper bound of the backoff interval based on the number of one-hop neighbors and number of transmission attempts. They prove that their algorithm reduces the number of collisions. Authors of [17] enable the nodes to change the upper and lower bounds based on the current network load and past history. In [18], authors introduce different subranges for backoff interval with respect to different network contention levels. Although these methods change their lower and upper bounds, final selection is done randomly within the new bound.

In this paper, we use BEB to tune the CW according to the current contention level. After tuning the CW, we introduce a new method of assigning backoff interval between . The individual parameters of each node like waiting time and queue length are taken into account to compute the backoff interval. These parameters help us to allocate a fair and effective medium access among the nodes. We ensure that unique backoff value is assigned to each node so as to avoid collision. Fuzzy logic is a simple and promising approach that extracts crisp and definite output from vague and ambiguous input parameters. Fuzzy logic has been widely used in wireless communication across various layers for computing, control, and decision making [19]. In [20] authors use fuzzy logic to calculate backoff interval to reduce contention over vehicular ad hoc networks. They control the current backoff interval using the past interval and success ratio of the node. Authors of [21] used fuzzy logic controller for early detection and prevention of congestion at the router buffer. They used delay rate and average queue length as input parameter and produced packet dropping probability as the crisp output. Authors of [22] have active router queue management based on conditions derived from Lyapunov stability theory. They used fuzzy congestion controller for the same.

In our method, each node collects input parameters from contending neighbors. The collected information is processed along with node’s own attributes and applied to membership functions to get fuzzy input parameters. By applying these fuzzy variables to the rules base, we can derive the backoff interval as crisp output.

#### 3. System Architecture

*Problem with IEEE 802.11*. In IEEE 802.11, the following steps are executed whenever a node wants to transmit a packet.(i)Node senses the medium.(ii)If the medium is idle for distributed interframe space (DIFS) period, then(a)the node generates a backoff counter randomly between ;(b)the backoff counter is decremented by one for every idle slot;(c)if the channel is busy, the backoff counter is paused until the next DIFS free period;(d)when the backoff counter reaches zero, the node starts transmission.

Each node uses BEB algorithm to find out the current contention window size (CW). The value of CW reflects the contention status of the channel. The minimum and maximum values of CW are called and with the default values of 31 and 1023, respectively. IEEE 802.11 updates CW as follows.(1)CW is initially set to .(2)After every unsuccessful transmission CW is doubled with the maximum limit of .(3)Upon a successful transmission, CW is reset to .

We note that the backoff value is randomly chosen between irrespective of the value of CW. Lower bound 0 changes the optimal order and frequency of channel access among nodes [16–18]. Previous studies have revealed that it greatly affects the average delay and throughput of the individual nodes. For larger number of nodes with heavy traffic, the number of collisions is more which leads to larger value of CW resulting in unfairness [23]. BEB never considers the traffic status (like waiting time or queue length) of the contending nodes for allocating the medium. The number of collisions can be reduced when the contending nodes are assigned with unique backoff value according to its traffic status.

*Proposed Design*. Our proposed algorithm enhances IEEE 802.11 DCF with additional level of contention resolution that prioritizes the contending nodes according to its queue length and waiting time. Each node learns about the contending nodes and computes a unique backoff interval between for itself. Contention window size (CW) is updated using IEEE 802.11 BEB. Each node needs to compute unique backoff interval by comparing its own data with the input parameters collected from contending nodes. In a dense network with larger nodes and heavy traffic, the collected input can be huge and vague. We make use of fuzzy logic at each node to find out its order for accessing medium.

Our system architecture is shown in Figure 1. It is clear that our fuzzy logic algorithm replaces random backoff selection and enhances IEEE 802.11. We specify that our algorithm can be incorporated along with the existing backoff algorithms [4, 6, 7, 9] too. In the next section, we explain our fuzzy logic algorithm in detail.