Soft Actor-Critic-Based Offloading and Resource Allocation Optimisation for Digital Twin-Aided Vehicular Networks

In this paper, we propose a multi-user multi-layer integrated sensing and communication (ISAC)-enabled digital twin (DT)-aided edge intelligent vehicular framework, which enables vehicular devices (VDs) to generate sensory and computational tasks and offload them to a roadside unit (RSU)-based MEC node for remote processing. Meanwhile, the DT technology pre-trains the collected real-time data and integrates with the updated data for processing, monitoring, and optimising the vehicular network in a virtual environment, and then improves the network resource utilisation efficiency. In addition, we formulate an optimisation problem to minimise the maximum service delay, including the transmission and computation delays, across all VDs. To solve this problem, we propose a soft actor-critic-based offloading and resource allocation optimisation algorithm, which jointly optimises computation task offloading decisions of all VDs, and communication and computation resource allocation. Simulation results show that our proposed algorithm outperforms the benchmarks in terms of service delay.