FROMS: A failure tolerant and mobility enabled multicast routing paradigm with reinforcement learning for WSNs

A growing class of wireless sensor network (WSN) applications require the use of sensed data inside the network at multiple, possibly mobile base stations. Standard WSN routing techniques that move data from multiple sources to a single, fixed base station are not applicable, motivating new solutions that efficiently achieve multicast and handle mobility. This paper explores in depth the requirements of this set of application scenarios and proposes Froms, a machine learning-based multicast routing paradigm. Its primary benefits are flexibility to optimize routing over a variety of properties such as route length, battery levels, ease of recovery after node failures, and native support for sink mobility. We provide theoretical, simulation and experimentation results supporting these claims, showing the benefits of Froms in terms of low routing overhead, extended network lifetime, and other key metrics for the WSN environment.