Visual Tracking for Teams of Miniature Robots

This paper describes an experimental visual tracking system for cameras mounted on miniature mobile, robots. The localization of the moving target robot is done ill two steps: (i) extracting boundary marks representing the target robot within the image, and (ii) reducing its projected position and orientation errors across images. Background segmentation is especially difficult in this problem, as there is no calibration with respect, to the scene from the mobile camera. Further difficulties arise with the constrained motion of mobile robots and performing the analysis in real-time. We also address all action selection scheme based oil error reduction as a solution for visual tracking from moving platforms. Experiments utilizing multiple miniature robots are presented to illustrate how this system can be used for following the leader, platooning, and target, interception behaviors.