Intelligent Offloading and Driving Strategy for Delay Minimization and Collision Avoidance in V2X Network

The task processing delay and road safety were the key challenges of vehicle-to-everything (V2X) network in sixth generation (6G) system. By offloading delay-intensive task of the vehicles to road side unit (RSU) and base station (BS), mobile edge computing (MEC) technology can reduce the processing delay of V2X network. However, it was difficult to dynamically associate between vehicles and MEC servers and offload the task for the moving vehicles, especially the low collision scenario. Thus, we considered the MEC-assisted multi-vehicle V2X system, where the vehicles can offload delay-intensive task to the multi-antenna RSUs and BS with the zero-forcing (ZF) receivers, and formulate the system delay minimization problem under the delay and collision constraints to satisfy the task processing delay and safety requirements in the V2X system. The multi-agent deep reinforcement learning (MADRL) algorithm was proposed to handle the non-convex problem by optimizing the offloading and driving strategy of vehicles. In this work, the reward function was designed bases on the task processing delay and vehicle collision, thus, the proposed MADRL algorithm can train vehicles to obtain the superior MEC server association, offloading ratio, and driving acceleration. Simulation results demonstrate that the system delay of the proposed MADRL algorithm can reduce by 16.78% and 4.55% compared to the Q-learning and full offloading schemes, respectively.