Inference-Based Distributed Channel Allocation in Wireless Sensor Networks

Abstract: Interference-aware resource allocation of time slots and frequency channels in single-antenna, halfduplex radio wireless sensor networks (WSN) is challenging. Devising distributed algorithms for such task further complicates the problem. This work studiesWSN joint time and frequency channel allocation for a given routing tree, such that: a) allocation is performed in a fully distributed way, i.e., information exchange is only performed among neighboring WSN terminals, within communication up to two hops, and b) detection of potential interfering terminals is simplified and can be practically realized. The algorithm imprints space, time, frequency and radio hardware constraints into a loopy factor graph and performs iterative message passing/ loopy belief propagation (BP) with randomized initial priors. Sufficient conditions for convergence to a valid solution are offered, for the first time in the literature, exploiting the structure of the proposed factor graph. Based on theoretical findings, modifications of BP are devised that i) accelerate convergence to a valid solution and ii) reduce computation cost. Simulations reveal promising throughput results of the proposed distributed algorithm, even though it utilizes simplified interfering terminals set detection. Future work could modify the constraints such that other disruptive wireless technologies (e.g., full-duplex radios or network coding) could be accommodated within the same inference framework.