DRL-based resource allocation for max–min fairness in V2V network

As vehicle-to-everything (V2X) standards continue to develop, ensuring both high throughput and fair resource allocation for safety-critical vehicle-to-vehicle (V2V) data has become a pivotal challenge in vehicular network design. To address this issue, this study proposes a resource allocation scheme that incorporates fairness constraints into the joint optimization of spectrum and power, thereby enhancing overall system throughput while maintaining fairness among V2V links. To solve the multi-objective optimization problem in dynamic vehicular environments, a deep reinforcement learning (DRL)-based framework is developed, which introduces improvements in experience sampling and reward modeling to enhance the stability and convergence efficiency of the learning process. Simulation results demonstrate that the proposed algorithm not only achieves optimized aggregate system throughput but also significantly enhances the minimum V2V link rate. Compared to baseline methods, the algorithm exhibits superior adaptability across varying vehicle densities, further validating its practical efficacy in dynamic scenarios.