Research on obstacle avoidance algorithm of self-driving bus based on driving risk assessment

This paper proposes an obstacle avoidance trajectory planning algorithm for autonomous driving buses, which enhances the vehicle’s ability to respond to dynamically changing environments. The proposed algorithm employs the lateral motion sampling method to generate multiple maneuver candidate trajectories, which are then evaluated using the normalized penalty function. Additionally, the algorithm considers the uncertainty of the future trajectory of obstacles and establishes a driving risk estimation model. The Monte Carlo sampling method is used to evaluate the driving risk of various maneuvering trajectories, and the algorithm considers collision probability, rear vehicle safety signs, and anxiety factors for the optimal candidate trajectory. Experimental results demonstrate that the proposed driving risk assessment method reduces the variance of the driving risk value by 83.14%, and the error value by 85.50% in the presence of perceptual uncertainty errors. The jump frequency of obstacle avoidance decisions can be reduced by 66.66%, and the variation range of the curvature is reduced by 32.29%. This approach generates a safe, comfortable, and stable obstacle avoidance trajectory in the presence of obstacles.