Safety-Aware Path Planning for Post-Disaster Rescue Vehicles

Emergency rescue vehicles play a vital role in dealing with traffic accidents and emergencies. Emergency rescue vehicles can reduce accident losses by arriving at the accident site in time. Reasonable planning of emergency vehicle routes can shorten the travel time of the route and improve the safety of the route. However, due to the dynamic and changeable actual road network environment and various uncertainties, congestion or accidents may occur at any time, resulting in low rescue efficiency and high risk of rescue routes. In response to the above problems, this thesis proposes a global path planning algorithm based on deep reinforcement learning. In response to complex and changeable road conditions, this thesis adds a path situation awareness module to the global path planning algorithm based on DQN, and proposes a Transformer-based prediction algorithm to perceive road condition changes, adjust the global path in real time, avoid congested paths and risky paths, improve the rescue efficiency of emergency rescue vehicles, and ensure driving safety. According to different evaluation indicators, this thesis conducts simulation experiments on the global path planning algorithm based on deep reinforcement learning. The experimental results show that the route travel time planned by the proposed algorithm is lower than that of the benchmark algorithm, the safety is higher than that of the benchmark algorithm, and the success rate of pathfinding has improved by 14% compared to the baseline algorithms.