To ensure the timeliness and smoothness of special vehicles performing emergency tasks in complex and congested urban traffic environment,a special vehicle route optimization strategy based on path search deep Q network(P-DQN)was pro-posed.The backtracking method was used to assist the deep Q network(DQN)to solve the problem of dead ends and loops in the path search process,and the artificial potential field mechanism was used to guide the DQN search path to avoid excessive path results.The roulette selection method and the greedy value adaptive adjustment mechanism were combined to further improve the accuracy of DQN when selecting road sections and suggesting driving speeds.Experiment simulated the real urban traffic on the InTAS dataset.The total value obtained using P-DQN is increased by about 16％compared with that using SOTA methods such as RERoute and CH.
维普
万方数据