Self-Driving Model for Safe Driving in High-Density Traffic Environment

In the safe driving of intelligent connected vehicles in high-density traffic environment, there are many problems to be solved, such as complex traffic situation processing, real-time decision response, and multi-vehicle interaction information capture. In order to improve driving safety and enhance decision robustness and adaptability, a safe driving model solution combining GNN and RL algorithm is proposed. By constructing the graph structure of the traffic scene to capture the interaction characteristics between vehicles and using TD3 algorithm to learn the optimal driving strategy, the intelligent motion planning for self-driving vehicle is provided. The simulation results show that the proposed model not only improves the training convergence speed, but also significantly improves the test success rate to the highest 98% under different traffic density and controls the collision rate and the deviation rate within 0.2% and 0.5%, respectively.