FROMS: Feedback Routing for Optimizing Multiple Sinks in WSN with Reinforcement Learning

In the domain of wireless sensor networks (WSNs), information routing is both a fundamental and challenging problem. In this work, we describe how information local to each node can be shared without overhead as feedback to neighboring nodes, enabling efficient routing to multiple sinks. Such a situation arises in WSNs with multiple, possibly mobile users collecting data from a monitored area. We formulate the problem as a reinforcement learning task, and apply Q-Routing techniques to derive a solution. Evaluation of the resulting FROMS protocol demonstrates its ability to significantly decrease the network overhead over existing approaches.