Risk-Aware Multi-Agent Advantage Actor-Critic Based Resource Allocation for C-V2X Communication in Cellular Networks

Intelligent transportation systems have emerged promisingly for industries to enable automated and safe driving. However, to satisfy reliability, environmental sustainability, and overall performance, deep reinforcement learning requires massive energy consumption with its computational demands. In this research, the risk-aware multi-agent advantage actor-critic (RA-MA-A2C)-based resource allocation (RA) is proposed for the cellular-vehicle-to-everything (C-V2X) network. The RA-MA-A2C considers collision risk when allocating resources such as frequency and power. By integrating risk assessment into the decision-making process, the RA-MA-A2C adjusts RA to mitigate collision risks and thereby increases the system’s effectiveness. The RA-MA-A2C’s performance is evaluated in terms of the success rate, completion time, vehicle-to-infrastructure link sum rate, and vehicle-to-vehicle links probability. The RA-MA-A2C demands 108 ms completion time with a 98.81% success rate, surpassing the performance of the existing offloading resource allocation-based deep reinforcement learning (ORAD) method.