QEBR: Q-Learning Based Routing Protocol for Energy Balance in Wireless Mesh Networks

For wireless mesh networks (WMN), energy balance has a significant impact on network lifetime. With the rapid growth of intelligent devices, learning from the environment and experience is a challenge for conventional routing protocols. To alleviate this problem, this paper proposes an energy balanced routing protocol based on Q-learning (QEBR) for WMN. Routing decisions are made by distributed reinforcement learning based routing models. A concept of neighbor energy sorting is proposed in QEBR to design the state and the reward function of Q-learning model. Simulation results demonstrate that QEBR has better performance of energy balance and packet loss rate, compared to the conventional method.