An Energy-Aware Q-Learning for Cluster-based Routing in WSNs

As sensor nodes are typically battery-powered or rely on energy harvested from the environment, recent works on the development and deployment of large-scale wireless sensor networks (WSNs) focus on reducing power consumption to improve network lifetime. Determining optimal routing for data to take the least energy path, or avoid non-responsive and energy-depleted nodes improves the lifespan of WSNs. Thus, we explore routing algorithms with computational intelligence that consider routing distance and residual energies of neighboring sensor nodes, while remaining energy-efficient. In this work, we present an energy-aware cluster-based routing protocol that leverages reinforcement learning. Our simulations show that using a simple Q-learning algorithm improves the energy efficiency of the network by more than 2x as compared to a traditional clustering algorithm. By incorporating energy awareness into the routing decisions, the operational lifetime of the network is improved by 1.4x. However, this leads to prioritizing routes with higher energy levels, which may not always be the shortest, resulting in a trade-off in energy efficiency. This work demonstrates the potential of reinforcement learning for developing energy-efficient routing protocols in wireless sensor networks.