Design of an Intelligent Vehicle Behavior Decision Algorithm Based on DGAIL

With the development of AI, intelligence level vehicles is increasing. Structured roads, as common and important traffic scenes, are most typical application scenarios for realizing autonomous driving. The driving behavior decision-making intelligent has always been a controversial difficult research topic. Currently, mainstream methods, which mainly based on rules, lack adaptability generalization to environment. Aimed at particularity vehicle decisions complexity environment, this thesis proposes an decision method DQN generative adversarial imitation learning (DGAIL) in structured road algorithm utilized GAIL generator. results show that DGAIL can preserve design reward value function, ensure effectiveness training, achieve safe efficient roads. experimental that, compared with A3C, GAIL, model spends less average training time 95% success rate straight scene merging scene, respectively. Apparently, effectively accelerate selection actions, reduce randomness actions during exploration, improve effect model.