Distributed cooperation in wireless sensor networks
The 10th International Conference on Autonomous Agents and Multiagent Systems - Volume 1
Decentralised reinforcement learning for energy-efficient scheduling in wireless sensor networks
International Journal of Communication Networks and Distributed Systems
| Hi-index | 0.00 |
In a multi-hop sensor network, idle listening is known to be a major source of energy drain. In this thesis we present the design and analysis of a novel Etiquette protocol to address this problem. Our key idea is to allow dynamic, flexible scheduling of inter-node communication, thereby minimizing energy wasted due to idle-listening. Scheduling communication is commonly used in single-hop networks for significant energy savings (e.g. GSM). However, scheduling communication is a nontrivial problem in a multi-hop network where packet load may not only vary in different parts of the network, but it may also fluctuate over time. The proposed Etiquette protocol allows nodes to schedule their communication in a completely distributed manner, while also allowing them the ability to change their communication schedule in response to the fluctuating packet load in the network. We present detailed simulation-based comparison of Etiquette protocol with S-MAC [52] and T-MAC [44]. Our simulation results show that Etiquette protocol reduces the average energy expended in delivering each bit by a factor of four as compared to S-MAC, and by a factor of more than two as compared to T-MAC. We show that Etiquette protocol allows a network designer to trade increased latency for reduction in energy consumption. The proposed protocol is simple and intuitive.