Multi-Agent Deep Reinforcement Learning Assisted Pre-connect Handover Management

This thesis proposes a MBB adopted handover mechanism, namely, pre-connect handover (PHO). PHO aims to provide a seamless and reliable handover for 5G networks. PHO utilizes DQN algorithm to facilitate the sequential decision-making problem of target base station (BS) selection based on the RSRQs and RSRQ change rates of all the surrounding candidate BSs. A MADRL solution is tailored to extend the DQN-assisted UE-associated PHO management for modeling a multi-UE scenario, where the autonomous agents learn the action policy by interacting with the environment in a distributed manner. The feasibility of PHO has been validated extensively via NS-3 and NS3-Gym. The experimental results demonstrated that the proposed PHO is not only achievable, but also that the DQN-assisted PHO technique can productively accomplish the optimal BS selection to maximize the PHO success rate. Moreover, the MADRL-assisted solution can also be conducted and effectively applied to a realistic multi-UE environment.