Vision-based Adaptive Tracking Control of a Mobile Robot: Algorithms and Experimental Validation

This paper mainly investigates the problem of vision-based tracking control for a mobile robot. A RGB-D camera is used to obtain the image with depth information. The proposed visual tracking algorithm, namely the multi-feature fusion Kernel Correlation Filters (MF-KCF), enhances the tracking capability by adding image features and dynamically adjusting scale factor and model updating strategy. To handle the issues with unknown velocity of the moving target and field-of-view for the camera, an adaptive controller is designed by incorporating the prescribed performance bound (PPB) technique. Finally, an experimental platform based on ROS (Robot Operating System) is established which contains mobile base, visual sensor and Jetson TK1, etc. The MF-KCF algorithm and PPB-based adaptive controller are successfully applied to the actual experiment to verify the effectiveness of the adaptive vision-based tracking system.