Autonomously Traversing Obstacles - Metrics for Path Planning of Reconfigurable Robots on Rough Terrain

The fixed chassis design of commonly employed mobile robots restricts their application to fairly flat environments, as the wheel diameters or the track heights impose hard limits on their mobility. Unstructured outdoor and urban environments alike comprehend many different invincible obstacles for most of those systems, like stairs, boulders or rubble. However, there are mobile robots with reconfigurable chassis providing a higher degree of mobility and enabling them to overcome such obstacles. Yet, current planning algorithms rarely exploit those enhanced capabilities, limiting these systems to the same environments as the fixed chassis robots. This paper focuses on the metrics used by our motion planner. The employment of a two-stage planning approach allows us to use different cost functions for the initial path search and the detailed motion planning step. The purpose of the initial search is to quickly find a fast environment-driven path to the goal. Hence, it uses fast computable heuristics to assess the drivability, i.e. a risk quantification and the utmost operation limits of the robot model. The detailed planning step determines the desired robot configurations. For this purpose, we consider the actuator controls, the system’s stability, an estimate of the traction, and the driving speed in addition to the quantities used in the first stage. We present experiments to illustrate the influence of the safety weights and real world experiments which prove the validity and feasibility of the metrics used by our motion planning algorithm.