Whole-Body Planning for Obstacle Traversal with Autonomous Mobile Ground Robots

Advances in the design of mobile robotics systems enable the application in new tasks like disaster response, inspection and logistics. Recently, autonomous robots have been a major focus of research. Compared to teleoperated machines, they work faster and more efficiently especially in environments with degraded connectivity. Without human supervision, the underlying algorithms need to be robust against unexpected circumstances to prevent damage to the robotic system and the environment. A common challenge for autonomous robots is the traversal of obstacles. To continue its task, the robot has to cross the obstacle without tipover instabilities. So far, research on prevention of vehicle tipover is mostly limited to simple systems with few degrees of freedom (DOF). In this thesis, a novel whole-body motion planning approach is proposed. By using a model of the world, the joint configuration is optimized for stability along a given path. The proposed method evaluates whether a safe traversal is possible and generates a motion plan that allows the robot to cross generic obstacles without tipover. Collisions are prevented by modeling them as constraints of the optimization. This approach is evaluated on a tracked vehicle with adjustable flippers and a five DOF manipulator arm. The proposed method leverages the flippers to improve stability by maximizing ground support and the arm to shift the center of mass. Additionally, the platform features various sensors to perceive its environment. Performance of the whole-body motion planning is evaluated in simulation and on the real robot. In multiple scenarios, it is shown that the approach effectively prevents tipover and increases robot stability.