Linear Visual Servoing-Based Control of the Position and Attitude of Omnidirectional Mobile Robots

This article proposes a Linear Visual Servoing (LVS)-based method of controlling the position and attitude of omnidirectional mobile robots. This article uses two markers to express their target position and attitude in binocular visual space coordinates, based on which new binocular visual space information which includes position and attitude angle information is defined. Binocular visual space information and the motion space of an omnidirectional mobile robot are linearly approximated, and, using the approximation matrix and the difference in the binocular visual space information between a target marker and a robot marker, the robot’s translational velocity and rotational velocity are generated. Since those are all generated based only on disparity information on an image, which is similar to how this is done in existing LVS, a camera angle is not required. Thus, the method is robust against calibration errors in camera angles, as is existing LVS. The effectiveness of the proposed method is confirmed by simulation.