Vision based navigation of heterogeneous robots

Vision-based navigation is one of the most attractive research subjects on mobile robot applications. An image-based navigation strategy was developed for a collaborative UAV-UGV system in this thesis. The work focuses primarily on terrain depth analysis, terrain image stitching, path planning, and motion control of the robot system. Novel algorithms for path planning and image stitching were developed as part of the thesis. In the terrain depth analysis, both stereo and structure from motion problems were tackled and their comparative advantages were highlighted. In addition, the performance of the onboard vision system used on UAV was evaluated using the system identification techniques to check its suitability for UAV motion control. In developing the new image stitching algorithm, inertial data was also used in addition to image data, and more accurate results were obtained than pure vision-based techniques. The developed path planning algorithm provided faster and more accurate results than the well-known A* and PRM, two widely used path planning algorithms. Simulation and experimental results obtained from the implementation of the algorithms show the effectiveness of the developed techniques for the collaborative navigation of heterogeneous robot systems