Enhancing Obstacle Avoidance in Dynamic Window Approach via Dynamic Obstacle Behavior Prediction

This paper proposes an enhanced local path planning method based on the Dynamic Window Approach (DWA), enabling a mobile robot to safely avoid obstacles and efficiently reach its destination. To overcome the limitations of the conventional DWA in handling dynamic obstacles and to improve goal reachability, the velocity term—originally evaluated solely by speed—was redefined as the distance difference between the robot’s predicted future position and the target destination. This modification allows the robot to more effectively anticipate its short-term position while simultaneously considering potential obstacle locations. In particular, a linear prediction model for dynamic obstacle behavior was introduced, which estimates the future positions of obstacles based on their current position, velocity, and heading direction. Under the assumption that obstacles maintain constant speed and direction over short intervals, this model enables the robot to proactively plan avoidance maneuvers before a collision risk arises. Furthermore, a novel risk assessment strategy was incorporated to enhance collision prevention. This approach categorizes obstacles in front of the robot according to both distance and angle, evaluates obstacle density in various directions, and guides the robot toward safer paths with fewer surrounding obstacles. The effectiveness of the proposed method was validated through extensive simulations, comparing the conventional DWA, a modified DWA with the new velocity term, and the proposed DWA with dynamic obstacle behavior prediction. The results demonstrated that the proposed approach significantly reduced the number of collisions and overall travel time, thereby confirming its superiority in highly dynamic and uncertain environments.