Research on intelligent lane-changing model based on multi-intelligent reinforcement learning

Most of the traditional automatic driving behaviours are based on mathematical formulae or vehicle dynamics and kinematics models, which ignore the driver's perception and decision-making ability in the actual driving process, and the generated trajectories are very different from the actual lane-changing trajectories. In this paper, we adopt an automatic driving system based on multi-intelligence reinforcement learning to optimise the lane-changing decision of the vehicle by combining multi-intelligence reinforcement learning algorithms and avoiding the interference of inter-vehicle information interaction. In this way, the self-driving vehicle can have the functions of complex environmental perception, path planning and behavioural decision making. Combined with the highway mixed-flow environment scenario, the vehicle lane-changing behaviour model based on the HMM model is carried out for comparative experiments, and the results show that the lane-changing intention model based on multi-intelligence body reinforcement learning performs better in lane-changing detection accuracy and can be adapted to traffic scenarios with different traffic flow densities.