Research on Dynamic Collision Avoidance for UGVs Using the Fusion of VO and APF Algorithms

Unmanned Ground Vehicles (UGVs), with their excellent maneuverability, have been widely utilized in various areas, including ground search and rescue, urban logistics, and public transportation. However, the limitations of the vehicle’s body structure pose challenges in sharp turning situations during the collision avoidance process of UGVs. The Artificial Potential Field (APF), as an obstacle avoidance algorithm featuring a smooth steering path, can effectively solve the problem of static collision avoidance. However, in dynamic scenarios, the APF algorithm may still produce paths that require sharp turns due to the lack of predictive ability. Therefore, this paper studies a novel algorithm (VO-APF) combining the Velocity Obstacle (VO) algorithm and the APF algorithm. Firstly, the Quaternion Vehicle Domain (QVD), considering relative position, is introduced into the VO algorithm to enhance its applicability in collision avoidance judgments within the vehicle domain. The reasonable point of collision avoidance is determined by predicting the collision avoidance time. Secondly, building upon the APF, the attraction function associated with collision avoidance targets and the repulsion function concerning obstacle vehicle areas are redefined. In the APF framework, UGVs are concurrently subjected to both attractive and repulsive forces. Finally, under the framework of modern traffic regulations governing collision avoidance, dynamic obstacle avoidance is achieved while minimizing intrusion into other vehicle domains, thereby enhancing overall safety during collision avoidance. The simulation results of three common dynamic collision avoidance scenarios show that the VO algorithm can fulfill the collision avoidance requirements, but its own vehicle intrudes into the obstacle vehicle domain for an extended period; The APF algorithm cannot meet the collision avoidance requirements in the crossing scenario, and it will also invade the vehicle domain of the obstacle vehicle; Compared to the APF and VO algorithms, the VO-APF algorithm reduces the vehicle domain overlap time by up to 47.53% and by at least 25.73%. It achieves shorter vehicle domain overlap times and improved safety performance, which effectively enhances the autonomous obstacle avoidance abilities of UGVs, providing a reliable guarantee for their autonomous navigation.