FLOW: An Efficient Forwarding Scheme to Mobile Sink in Wireless Sensor Networks

We propose a learning-based approach to efficiently forward data to a mobile sink in a wireless sensor network. Specifically we assume a push application where the mobile sink does not initiate the query. Furthermore, the sink moves in a randomized pattern within the sensor field. In the presented scheme, Moles (nodes that sense the sink) learn the sink’s movement pattern over time and statistically characterize it as probability distribution functions. This information collected at the moles is used in a distributed fashion to calculate the likelihood of a node being on a good path to the sink which is used to determine whether to forward data through the node.