6-1 Introduction
The previous chapter assessed the statistical characteristics of the Carrier Sensing Multiple Access with Collision Avoidance (CSMA-CA) protocols over multihop networks using the IEEE 802.15.4 CSMA-CA beaconless protocol as an example [16]. Although this protocol has been specified especially for WSNs, the results of chapter 5 highlighted the performance limitations of this protocol over multihop networks, i.e., it has been shown that the average end-to-end delay and probability of lost packets increase significantly with increasing number of hops, average number of nodes contending to access the channel, or the traffic intensity. The deep investigation conducted in chapter 5 showed that this degradation is attributed mainly to the heavy tail of the probability distribution of the medium access delay.
The IEEE 802.15.4 CSMA-CA protocol, like other existing CSMA-CA protocols, consists of two fundamental mechanisms: channel assessment and back-off. The channel assessment process is used to prevent a node from commencing transmission while the channel is occupied with transmissions from other nodes. The back-off algorithm is used to regulate channel access by making a node wait for a random period of time after a collision or busy channel assessment. Therefore, the efficiency of the CSMA-CA protocol depends mainly on the probability distribution used to generate the back-off periods.
Most existing CSMA-CA protocols employ Binary Exponential Back-off (BEB). In these schemes, back-off durations are generated according to a uniform distribution. The window of this distribution is set to an initial value at the start of contention to access the channel and is reset to the same initial value after each successful transmission or when a packet is dropped. The window is doubled whenever a collision or busy channel assessment occurs. Thus, the windows of back-off stages in BEB increase multiplicatively based on the status of the channel (either busy channel or collision) which means that the process that is yielded from BEB is a power law process. From a statistical perspective, the probability distribution of a power law process is a heavy tailed distribution and hence we can conclude that the BEB scheme of CSMA-CA protocol is the main cause of the performance degradation of these protocols over multihop WSNs.
M. Baz, PhD Thesis, University of York 2014
the back-off periods of a node according to its distinct traffic demands. In a typical multihop network, the traffic pattern is heterogeneous since each node has its own traffic characteristics that differ substantially from other nodes and hence each node needs to contend to access the channel differently. Comparing this requirement with the BEB scheme demonstrates that in BEB, all nodes execute the same procedures irrespective of their traffic loads. The second main limitation of the BEB scheme is that this scheme does not provide a mechanism to resolve the cause of unsuccessful transmission. BEB doubles the back-off windows blindly without providing remedial action for the underlying causes of failed transmission. An illustration for this shortcoming can be acquired by considering the two BEB mechanisms of two standards, the IEEE 802.11 [97] and IEEE 802.15.4 [16]. In the IEEE 802.11 the back-off window is doubled after each collision regardless of whether this collision was due to a non-hidden or hidden node. Conversely, in the IEEE 802.15.4 standard, the back-off window is doubled after a busy channel assessment.
Although a number of ideas have been proposed to improve the performance of BEB, most of them are not specifically designed for multihop networks which make them inadequate in achieving acceptable performance readings for these networks in terms of network throughput, node lifetime and packet delay. Hence there is a high demand to develop a CSMA-CA mechanism for multihop networks. The importance of this chapter for the remaining of the thesis is that it provides the back-off scheme required to make the MAC protocol effective for multihop networks. Upon this proposed scheme other enhancements, e.g., the duty cycle management scheme, routing protocols that are developed in the remaining chapters are incorporated. This chapter proposes a novel CSMA-CA scheme for multihop networks. The proposed algorithm allows each node to adjust the parameters of the probability distribution used to generate the back-off intervals. This aim of this tuning is to achieve the following objectives: reducing the end-to-end delay, increasing throughput and minimising the energy consumption. End-to-end delay is reduced by letting a node to assess the channel with a high probability when it is presumably idle, and back-off with a high probability when the channel is potentially busy, hence a node can avoid deferring access to the channel for the additional random period. The collision probability is reduced by employing a novel collision resolution algorithm that accounts for both non-hidden and hidden node collisions. Reduction of the time required to service a packet and the probability of collision prolong the lifetime of nodes by reducing the energy consumed in contention to access the channel. The proposed protocol employs the Gamma distribution [182] to generate the back-off intervals. The Gamma distribution has been selected because it is a highly parametric distribution which provides an excellent basis for adapting to the wide variety of potential traffic distributions. For example, the exponential, Poisson, Erlang, Chi-square and normal distributions can be derived from a Gamma distribution simply by adjusting its parameters [182]. Another great benefit of employing the Gamma distribution in the CSMA-CA protocol is to mitigate the high level of variation found in the BEB protocols. Employing the Gamma distribution generates a Gamma process that has a low level of variation compared to that yielded
M. Baz, PhD Thesis, University of York 2014
from the power law process in the BEB scheme. Importantly, the protocol is lightweight and can be applied to any wireless network, as it does not require modification of the physical layer standards. The simulated outcomes demonstrate the bountiful benefits of the proposed protocols.
The rest of this chapter is organised as follows. Section 6-2 provides an overview of the binary exponential back-off process. Section 6-3 explores the related work in order to highlight their contributions and limitations and to motivate the proposed work. Section 6-4 discusses the statistical characteristics of the Gamma distributions and section 6-5 exploits this discussion to highlight the underlying approach of the proposed protocol. Section 6-6 provides the pseudo-code of the proposed protocol, section 6-7 assesses the benefit of the proposed protocol and finally section 6-8 concludes this chapter.