Efficient Node Discovery in Mobile Wireless Sensor Networks

Energy is one of the most crucial aspects in real deployments of mobile sensor networks. As a result of scarce resources, the duration of most real deployments can be limited to just several days, or demands considerable maintenance efforts (e.g., in terms of battery substitution). A large portion of the energy of sensor applications is spent in node discovery as nodes need to periodically advertise their presence and be awake to discover other nodes for data exchange. The optimization of energy consumption, which is generally a hard task in fixed sensor networks, is even harder in mobile sensor networks, where the neighbouring nodes change over time.

In this paper we propose an algorithm for energy efficient node discovery in sparsely connected mobile wireless sensor networks. The work takes advantage of the fact that nodes have temporal patterns of encounters and exploits these patterns to drive the duty cycling. Duty cycling is seen as a sampling process and is formulated as an optimization problem. We have used reinforcement learning techniques to detect and dynamically change the times at which a node should be awake as it is likely to encounter other nodes. We have evaluated our work using real human mobility traces, and the paper presents the performance of the protocol in this context.