Vision-based mobile robot speed control using a probabilistic occupancy map

This paper describes a method of controlling robot speed using a probabilistic occupancy map. It is usually necessary for a robot to make sure that a target region is free before entering it. If the robot is not confident in the state (free or occupied) of the region, the robot has to make enough observations to be confident. If the distance to the region is long, the robot can expect to have enough observations until reaching it. If the distance is short, however, the robot may have to slow down to make enough observations. Based on this consideration, we develop a method for controlling robot speed by considering the state of a target region. The method is applied to our mobile robot with omnidirectional stereo and a laser range finder. The robot successfully moved around in an unknown environment by adaptively controlling its speed.