Visual Safety of Robot Networks based on Control Barrier Functions with Spherical Obstacles

This paper addresses the target pursuit task by multiple mobile robots equipped with visual sensors in an environment with spherical obstacles. First, we define a target as visible to a robot when feature points on the target are inside the robot's field of view and are not occluded by an obstacle. To formulate this visibility, we derive safe sets for maintaining the field of view and for avoiding occlusion. Next, in order to achieve visibility of the robot network, we propose two methods for online visibility determination based on the mixed integer programming and the dynamic average consensus algorithm. Finally, the effectiveness of the proposed method is verified by simulation.