Pheromone-based In-Network Processing for Wireless Sensor Network Monitoring Systems

Abstract

Monitoring spatio-temporal continuous fields using wireless sensor networks has emerged as a novel and efficient solution. The development of energy efficient query dissemination and data collection algorithms for environments where only a small subset of nodes has relevant readings is a challenging problem if no information about the location of these nodes is available. Monitoring these data requires not only an initial discovery stage but also a continuous search for new relevant data due to field variations in time. One solution to this problem is to let nodes cooperate and decide jointly which data are relevant and their locations. Trails to relevant data can be distributively generated as insect colonies do using pheromone-based indirect communication. In this work, we propose PhINP, a probabilistic Pheromone-based In-Network Processing mechanism to monitor information using WSNs. Our proposal takes advantage of both a pheromone-based iterative strategy to direct queries towards nodes with relevant information and query- and response-based in-network filtering strategies to reduce the quantity of nodes selected to answer the query. Additionally, nodes use reinforcement learning to improve the routing performance of queries. We demonstrate by extensive simulations that using PhINP mechanism the query routing cost can be reduced by approximately 60% over flooding, applying the same in-network filtering strategy, with an error below 1%.