Decentralised Adaptive Sampling of Wireless Sensor Networks

Wireless sensor networks are being deployed in an increasing number and variety of applications. Currently, most of these systems adopt a centralised control mechanism, but this has issues associated with scalability, robustness, and dynamism that often exist in such networks. Given this, decentralised approaches are appealing. However, the design of an efficient decentralised control regime is difficult as it introduces additional control issues related to the interactions between the network’s interconnected nodes given the absence of a central coordinator. Within this context, we consider how such approaches can be applied to the problem of ada ptive sampling in energy-constrained networks. In particular, we represent each sensor as an autonomous agent and develop a decentralised algorithm for a deployed sensor network in the domain of flood monitoring. This algorithm is then empirically shown to perform significantly better than a non-adaptive benchmark.