Target Importance-Based Obstacle Avoidance Path Planning for Autonomous Driving

Conventional path planning methods primarily rely on the perceived location and speed of target obstacles, the importance of these obstacles is rarely considered, e.g., the importance of large truck and small one for path planning is obviously different. To solve this problem, an autonomous driving path planning method based on target importance is proposed. Different from the traditional path planning method, the reward function is modified to be an combination of the vehicle type and distance from obstacles, the background vehicles are divided into large vehicles and small ones, and different reward function weights are set, based on which sampling dynamic planning is performed for path planning. The proposed algorithm is tested on NGSIM and its performance is compared with conventional path planning methods. The cumulative reward obtained by the proposed method is significantly higher than that obtained by the conventional method, indicating that our method is more reliable for obstacle avoidance of large vehicles.