A practical heterogeneous network optimisation algorithm based on reinforcement learning

Abstract

Owing to the increasing and diversifying service requirements of wireless communications, wireless networks must coexist with heterogeneous radio systems. To realise the interconnection between different networks, it is important for the radio access network elements, such as the cellular network base stations (BSs) and the wireless local area network (WLAN) access points (APs) to be reconfigurable based on the real-time network environment. In this paper, we propose an efficient distributed reconfiguration algorithm for heterogeneous networks: the dynamic network self-optimisation algorithm (DNSA). This algorithm is based on the Q-learning algorithm and the self-optimisation of each network entity acting as independent agents. In the proposed algorithm, multiple agents perform the optimisation cooperatively to reduce the system blocking rate and improve network revenue. The dynamic network self-optimisation problem is transformed into a multiple-agent reinforcement learning problem which has much lower complexity and better performance.