Development and Realization of an autonomous driving concept for planetary vehicles

Abstract

Challenging applications are present for the �eld of robotics even outside the laboratory conditions in the open environment. To enable the commercial exploration of the robotics, `autonomous navigation' of a robotic vehicle plays a critical role. In the recent past, there have been significant progress made towards the supervised autonomous robotic capability for remotely controlled scientific exploration of the planetary surfaces. This thesis investigates the existing driving concepts for the autonomous vehicles with emphasis on planetary applications. Navigating through an unknown environment the robot faces challenges like accurate obstacle detection and positioning of the robot. During such navigation there might be some situations, where the obstacles are of the same texture and color of the ground surface. Distinguishing the obstacles with respect to the ground surface thus becomes very difficult. To overcome this problem an approach of using three dimensional coordinates of the points is used e�ectively. Thus the thesis work focuses on use of three dimensional coordinates of the points to detect an obstacle. Some of the innovative methods used for the obstacle detection have been explained in this thesis. The positioning of the robot is done with respect to a starting point using `camera motion estimation'. This camera motion estimation is part of the `stereo vision system'. For this study work only one sensor is used i.e. stereo vision system. The focus of this thesis is on the navigation strategy based on the vision system. The various stages involved in developing this navigation strategy are explained. The necessary strategy with the related software coding have been implemented and the real time experiments are also carried out. The results of the various experiments carried out with interesting situations are also analyzed. Various failures have been observed during these experiments and the possible solutions for the future development are suggested. The developed navigation strategy is based on two basic layers, `obstacle check' and `goal check'. To incorporate any future developmental work, like addition of extra layers in the current architecture is also discussed.