Collision avoidance method for multiple autonomous mobile agents by implicit cooperation

This paper proposes a collision avoidance method, called the cooperative collision avoidance method, for an environment consisting of multiple autonomous mobile agents. We also consider tasks such as map exploration and object collection, in addition to the collision avoidance. The method is based on local measurements and local planning at every sampling period in real-time. The proposed method is an extension of the conventional velocity obstacle (VO) method which provides collision detection among moving agents. The proposed method utilizes the uniformity of agents enabling the agents to share information of collisions without explicit communication. Sharing collision information allows agents to cooperate implicitly with each other to avoid collision by using the common VO approach. Furthermore, we introduce a velocity index which evaluates velocities to achieve good task performance and collision avoidance simultaneously. Finally, the efficiency of the cooperative collision avoidance method is demonstrated by analyses and simulations on examples.