A Study on Channel Polling Mechanisms for Wireless Sensor Networks

Abstract

Due to the highly application specific nature of WSN, hundreds of Media Access Control (MAC) protocols have been proposed in the past. The focus of these protocols has been on optimizing the performance parameters such as energy, delay, throughput and reliability. Among these, improving energy efficiency of WSN has always been the primary goal. MAC protocols have taken various approaches to manage the energy consumption of WSN nodes efficiently, channel polling using Low Power Listening (LPL) is one of them.

Channel polling refers to the channel listening by the receiving nodes mostly in asynchronous MAC protocols. The channel polling process significantly governs energy, delay and lifetime of the network. Therefore, it is required to adjust the polling intervals in accordance with the incoming traffic patterns. The previously proposed protocols either polled the channel periodically regardless of the traffic arrival patterns, or attempted to adjust the polling intervals based on the history of past idle & busy polls. However, none of the previous protocols focused on studying the influence of using probabilistic distributions for selecting polling interval distributions on the performance of MAC protocols in WSN. For example, Boost-MAC and AXMAC decide polling intervals based on the history of previous polls, whereas AS-MAC sets the polling intervals based on the analysis of neighbor schedules. Hence, the research gap identified for this dissertation is to study the influence of varying polling interval distributions for the MAC protocol and to develop an asynchronous duty-cycle based MAC protocol: Adaptive and Dynamic Polling-MAC (ADP-MAC), which is based on this concept.

The protocol ADP-MAC has been developed to be dynamic and energy efficient which could adapt to a wide range of WSN applications, particularly those with dynamic traffic patterns. The study took the novel approach of switching the polling interval distribution of the receiver nodes by monitoring the Co-efficient of Variation (CV) of the incoming traffic. The CV is hence intended to be used as an indicator of the variance present in not just the generated traffic but ultimately the traffic arriving at the receiving nodes. A Receiver Initiated Pseudo-Synchronization (RIPS) scheme has been integrated with the protocol to ensure that nodes get back into communication after a certain period of being out of synchronization. To represent different applications of WSN, Constant-Bit Rate (CBR), Poisson and Bursty Arrivals have been used; whereas three types of polling distributions have been studied: Deterministic, Exponential and Dynamic. The performance of ADP-MAC has been evaluated with the features of Dynamic Polling set ON and OFF; when the Dynamic Polling is set OFF, the protocol conducts either deterministic or Exponential Polling. Closely linked to the energy performance of ADP-MAC is the use of packet concatenation and block acknowledgment mechanisms, which have also been analyzed.

The single-hop and multi-hop performance of ADP-MAC has been evaluated through implementing the design in TinyOS for mica2 platform. Avrora emulator has been used for simulations and Awk script was written for data parsing from the output file. The experiments are conducted for varying number of nodes, message generation intervals and polling intervals, among other parameters. The mechanisms of Adaptive & Dynamic Polling and packet concatenation are separately studied to unveil the contribution of each aspect of ADP-MAC.

It has been found that in terms of energy consumption and delay, ADP-MAC with Deterministic Polling serves best for CBR Arrivals, ADP-MAC with Exponential Polling serves best for Poisson Arrivals and finally, ADP-MAC with Dynamic Polling serves best for the Bursty Arrivals. No trade-off between the delay and energy has been seen for ADP-MAC because when there is a better match between the instants of packet generation at the source node and channel polling at the receiver node, the preamble transmissions reduce resulting in both the reduction of energy consumption and delay at the same time. These observations have led to the major finding of this dissertation that when the traffic arrival and polling interval distribution of ADP-MAC are in conformance, the performance in terms of both delay and energy turns out to be the best. On the other hand, ADP-MAC with Dynamic Polling results in best performance in terms of packet loss regardless of the type of arrivals.

The performance comparison of ADP-MAC has been shown against an established MAC protocol: Synchronized Channel Polling MAC (SCP-MAC). SCP-MAC has been chosen as a benchmark protocol because it combines the properties of synchronous (loose synchronization) & asynchronous protocols (channel polling). Although SCP-MAC is defined as synchronous, it performs additional adaptive polls after each successful reception. We developed a research gap to evaluate the performance of dynamic selection of polling interval distributions instead of performing adaptive polls as in SCP-MAC. The results show that ADP-MAC outperforms SCPMAC in terms of energy consumption, delay and packet loss. The major reason for the superior performance of ADP-MAC is that it does not perform explicit synchronization and reduces contention, idle listening and overhearing. Moreover, the Adaptive Polling mechanism of ADPMAC dynamically selects the best polling interval distribution based on traffic arrival patterns, hence reducing the energy consumption associated with preamble transmissions and excessive polling.