Local Minimum Avoidance in Potential Field Method Through Virtual-Wall Generation and Wall Following

This paper proposes a novel method to solve the local minimum problem of conventional potential field methods by generating virtual walls to achieve autonomous movement in unknown environments. The existence of local minimum can be attributed to two main factors: (i) the robot enters a concave region and becomes trapped, and (ii) the attractive and repulsive forces balance out when an obstacle is located directly in the direction of the goal. The proposed method addresses these challenges by generating a virtual wall along obstacles with concave shapes, preventing the robot from entering such regions. Furthermore, when a local minimum occurs owing to the force equilibrium caused by obstacles in the goal direction, the robot executes a wall-following behavior to escape from the local minimum. We confirmed that the proposed method enables a robot to move toward the goal more efficiently than the conventional method by generating virtual walls and following them in an environment with local minimum.