Intelligent Handover Orchestration in Beyond 5G and Urban V2X Dual Connectivity Networks: A Deep Reinforcement Learning Approach

초록

The coexistence of cellular (Uu) and sidelink (PC5) interfaces in 5G Vehicle-to-Everything (V2X) communications introduces complex handover management challenges, particularly in urban environments with heterogeneous infrastructure comprising macro base stations and Road-Side Units (RSUs). This paper presents a Double Deep Q-Network (DDQN)-based dual connectivity controller that intelligently orchestrates handover decisions and interface selection while leveraging RSU collaboration for enhanced V2X communications. The proposed solution introduces a comprehensive 10-dimensional state representation explicitly incorporating RSU proximity, RSRP, SINR, cluster dynamics, and handover history, enabling proactive handover decisions. A sophisticated multi-objective reward function balances handover frequency against performance gains through explicit handover cost penalties and ping-pong detection mechanisms. RSUs serve as local orchestrators, predicting optimal handover timing based on vehicle trajectories and cluster formations. Extensive simulations demonstrate that the DDQN controller achieves 94.5% PDR, 13.6 ms average latency, and 0.694 interface diversity, while reducing the ping-pong ratio to only 5%. The system demonstrates real-time feasibility with only 1.5% processing overhead and 323.8 ms average step time in 50-vehicle scenarios, proving balanced performance across multiple objectives. © 2013 IEEE.

키워드