DRL-Based Two-Timescale Resource Slicing Scheme in V2X Communications

Network slicing has emerged as a key technology to satisfy the diverse quality of service (QoS) requirements of heterogeneous vehicle-to-everything (V2X) services, such as enhanced mobile broadband (eMBB) and ultra-reliable low-latency communications (URLLC). However, in dynamic V2X scenarios, the time-varying characteristics of vehicle density, network topology, and service demands pose significant challenges to efficient resource allocation schemes. Consequently, traditional static strategies are prone to resource fragmentation, low spectrum efficiency, and inability to meet strict and dynamic service level agreement (SLA) requirements. To tackle these challenges, we propose a two-timescale resource allocation framework for V2X communications based on network slicing, which aims to maximize the system QoS satisfaction of heterogeneous services. Meanwhile, we develop a two-timescale algorithm framework by integrating twin delayed deep deterministic policy gradient (TD3) and genetic algorithm (GA), thus realizing a complete cross-timescale optimization. Simulation results show that both algorithms in the two-layer framework exhibit satisfactory convergence performance and the proposed framework outperforms baseline schemes in improving system QoS satisfaction.