Dynamic obstacle avoidance for unmanned vehicles based on motion prediction and improved DWA algorithm

To address the issue of inefficiency in unmanned vehicle obstacle avoidance caused by the difficulty in predicting dynamic obstacle trajectories in complex environments. This paper investigates a dynamic obstacle speed detection method based on 2D LiDAR and Kalman filtering, with unmanned vehicles as the research subject. A kinematic model of the unmanned vehicle is developed, and an improved Dynamic Window Approach (DWA) is proposed. The simulation results indicate that, compared to the traditional obstacle avoidance algorithm, the improved DWA algorithm achieves a 13.7% reduction in traveling time and a 12.5% decrease in traveling path length when addressing slow-moving obstacles. For fast-moving obstacles, the algorithm reduces traveling time by 14.1% and shortens the traveling path length by 15.7%. Additionally, a dynamic obstacle avoidance experimental platform for unmanned vehicles is developed to conduct verification experiments. The experimental results show a 3% error compared to the simulation, confirming the effectiveness of the proposed method.