Q-Learning Algorithm Enabled Topology Control Scheme in Power Line Communication Networks

Abstract

The topology control technology is of great importance in the Power Line Communications (PLC) networks for smart power grid. Due to the weak topology controlling ability of current large-scale PLC networks, the throughput of current PLC networks is limited seriously and fail to support more real-time services. To deal with this problem, this paper proposes a Q-Learning algorithm enabled topology control scheme of PLC networks. This robust topology control scheme introduces the Q-learning algorithm into the CSMA/CA protocol and adopts the Markov process to build the networking model of PLC system. Through period on-line learning by proposed algorithm, the robust topology can be established in PLC networks. Test results show that the proposed approach can improve topology control ability for real-time services by smart grid IoT (SG-IoT) systems, in terms of both packet loss rate and time delay performance with great feasibility and efficiency.