Partner selection in self-organised wireless sensor networks for opportunistic energy negotiation: A multi-armed bandit based approach

Abstract The proliferation of “Things” over a network creates the Internet of Things (IoT), where sensors integrate to collect data from the environment over long periods of time. The growth of IoT applications will inevitably involve co-locating multiple wireless sensor networks, each serving different applications with, possibly, different needs and constraints. Since energy is scarce in sensor nodes equipped with non-rechargeable batteries, energy harvesting technologies have been the focus of research in recent years. However, new problems arise as a result of their wide spatio-temporal variation. Such a shortcoming can be avoided if co-located networks cooperate with each other and share their available energy. Due to their unique characteristics and different owners, recently, we proposed a negotiation approach to deal with conflict of preferences. Unfortunately, negotiation can be impractical with a large number of participants, especially in an open environment. Given this, we introduce a new partner selection technique based on multi-armed bandits (MAB), that enables each node to learn the strategy that optimises its energy resources in the long term. Our results show that the proposed solution allows networks to repeatedly learn the current best energy partner in a dynamic environment. The performance of such a technique is evaluated through simulation and shows that a network can achieve an efficiency of 72% against the optimal strategy in the most challenging scenario studied in this work.