5-1 Introduction
The first main objective of this thesis is to design an effective MAC protocol for multihop WSNs that can achieve three concurrent goals: increasing the throughput of a network, minimising the end-to-end delay of packets and prolonging the lifetime of nodes. The discussion presented in chapter 3 demonstrates that the Carrier Sensing Multiple Access with Collision Avoidance (CSMA-CA) protocols as decentralised schemes with low overheads are suitable for WSNs. Moreover, it has been shown that the design spaces of most of the current CSMA-CA protocol are inadequate to accommodate the unique characteristics of multihop networks, e.g., heterogeneous traffic patterns and distinct channel conditions for each transmitter-receiver pairs. In order to reveal the main limitations of the current CSMA-CA protocols, and thereby to overcome them during the development of the proposed protocol, there is a need to acquire a deep and accurate understanding for the behaviour of CSMA-CA under different operational conditions.
An exact and complete understanding of the CSMA-CA protocol can be acquired through statistical analysis as such analysis can abstract the stochastic behaviour of the protocol as a probability distribution. Therefore all the possible states of the system can be predicted readily and moreover the design of CSMA-CA protocols can be accomplished without missing information. Another great advantage of the statistical analysis is its ability to quantify the relationships between hidden variables that affect the performance of the CSMA-CA protocol and that cannot be derived from other analytical or simulated models: e.g., the effect of variation of the back-off on the channel assessment [6].
In general, analysing the CSMA-CA protocol statistically is a challenge, and this challenge increases tremendously when the analysis targets multihop WSNs as these networks constitute a highly random system. Explanations for the randomness of a typical multihop WSN can be drawn by considering the following facts. Firstly, the lack of a de-facto topology that can represent wide deployments of multihop WSNs, and thus a typical representation of a multihop WSN is a random planar process. Secondly, the heterogeneity of the traffic load of nodes, as in a multihop WSN the traffic intensity of a node comprises the traffic generated by the node itself as well as the traffic forwarded by other nodes for relaying. Thus, the traffic intensity of a node is a time-spatial random process. Finally, as a typical multihop WSN is a large-scale
M. Baz, PhD Thesis, University of York 2014
network in which each node has a different number of neighbours and each node has its own traffic intensity, the view of the channel conditions of each transmitter– receiver pair differs substantially from others. Moreover due to the low-data rate of WSNs, a packet encounters a high amount of both queuing and medium access delay, thus the contention between nodes becomes subject to the randomness of the topology, traffic and service time. Accounting for these factors in analysing the statistical characteristics of CSMA-CA protocols over a multihop WSN requires a new paradigm that is able to reflect these different levels of randomness.
This chapter proposes a novel approach to analyse the statistical characteristics of CSMA-CA protocols over multihop WSNs [6]. The proposed approach represents the multihop WSNs as a three layered model: a topology model, a routing model and a queuing model. The topology model abstracts the underlying structure of the network using set theory to describe the physical relations between nodes, e.g., transmission range and carrier-sensing range. The routing model quantifies the traffic load of each node based on the originator-final destination traffic rates and the multipath routing policy. The queuing model represents a node as a GI/G/1 queue [139] in which the inter-arrival distribution is constructed from the routing model. The service time distribution of a queue is derived from the characteristics of the CSMA-CA protocol considering the physical constraints from the topology model; this model then uses queuing theory to obtain the other quantities, including the distributions of inter-departure and waiting times. The benefit of the proposed modelling approach is threefold: firstly, it facilitates modelling a multihop network as a network of queues considering the effects of both the physical and routing layers; secondly it enables assessment of the performance of a MAC under a wide variety of situations just by modifying the corresponding models without requiring reconstruction of the entire analytical framework; and finally, it presents a general framework that can be easily adapted for use with any other medium access control protocols (either contention based or scheduled based) by varying the service time distribution. Using the proposed framework, the Moment Generating Functions (MGFs) of the probability distributions of the MAC service time, inter-departure, and waiting times are obtained. From these distributions, comprehensive assessment of the behaviour of the CSMA-CA protocol is derived including different performance metrics (average end-to-end delay, probability of dropping a packet and throughput). In order to demonstrate the accuracy of the proposed analytical framework, we use it to analyse the IEEE 802.15.4 CSMA-CA protocol and then compare its results with simulation outcomes. The main aim of this comparison is not only to validate the accuracy of the proposed model but also to highlight the main limitations of the existing CSMA-CA protocol and enable us to overcome them when developing the proposed MAC protocol. Moreover, this framework constitutes the building block for the other analytical model presented in the subsequent chapters.
The remainder of the chapter is organised as follows: in section 5-2 the key existing analytical models and their limitations are summarised, section 5-3 overviews the statistical theorem used in deriving the analytical model and section 5-4 presents the proposed analytical framework. Section 5-5 uses the proposed analytical model to
M. Baz, PhD Thesis, University of York 2014
evaluate the performance of the IEEE 802.15.4 CSMA-CA protocol and Section 5-6 presents and discusses the results. Finally, section 5-7 concludes the chapter.