DRQN artificial intelligence algorithm and CGAN analysis for intelligent driving vehicle path design

In response to the low efficiency and poor reliability of traditional intelligent driving vehicle path autonomous planning algorithms, a new path autonomous planning algorithm is proposed on the basis of deep recurrent Q-Network algorithm and conditional generative adversarial network as the core. The experimental results showed that in the simulated environment, the average accuracy before and after training was 0.79 and 0.91, respectively. After utilizing conditional generative adversarial networks for further optimization, the two were 0.83 and 0.95, respectively. The average success rate, running time, and reward value of the algorithm were 88.20%, 15.94s, and 0.81, respectively. After optimization, the three were 92.10%, 11.97s, and 0.90, respectively. In practical applications, the average success rate and absolute deviation of the proposed algorithm were 86.52% and 1.85%, respectively. After further optimization, the two were 91.16% and 1.56%, respectively. The average distance and absolute deviation between the algorithm and obstacles were 3.37m and 0.30m, respectively. After optimization, the two were 3.75m and 0.31m, respectively. The average accuracy of the algorithm was 0.82 in different traffic environments. After optimization, the average accuracy was 0.86. The research can optimize the efficiency of autonomous path planning and realize the application of intelligent driving vehicles in more fields.