Motion Planning for Minimizing Wheel Wear in Planetary Rovers on Rough Terrain

Kurzfassung

Planetary rovers operating on abrasive terrain experience progressive wheel degradation, with steering maneuvers on rough surfaces identified as a significant contributing factor. While existing motion planners optimize for geometric feasibility, collision avoidance, and energy efficiency, they typically treat all terrain as mechanically equivalent and do not account for the physical consequences of steering maneuvers on wheel lifespan.

This thesis investigates motion planning strategies that reduce steering-induced wheel wear by incorporating terrain-aware cost formulations into the planning process. A curvature-based cost model is developed that penalizes high-curvature maneuvers in proportion to local terrain roughness. Path curvature and its rate of change serve as computationally efficient proxies for steering effort, capturing the kinematic relationship between rover motion and wheel-level steering demands. This formulation shifts wear mitigation from the execution layer, where operational strategies such as modified driving primitives have previously been applied, to the deliberative planning layer.

The terrain-aware cost is integrated into two planning frameworks: a sampling-based kinodynamic RRT* planner with nonlinear trajectory optimization, and a search-based lattice A* planner extended from the Nav2 SMAC framework. For the sampling-based planner, a hybrid rejection–corridor sampling strategy is developed that improves steering feasibility and accelerates convergence; this is followed by a smoothing stage that incorporates terrain-aware curvature penalties. For the lattice-based planner, terrain-dependent edge costs are introduced that penalize high-curvature motion primitives on rough terrain.

Both planners are evaluated in simulation using the Lightweight Rover Unit (LRU) platform across hand-crafted navigation scenarios and 100 randomly generated environments. Results demonstrate that terrain-aware costs generally reduce wheel-level curvature and cumulative steering effort in rough terrain regions, though the mechanism differs by planner. The lattice-based planner achieves consistent improvements by selecting alternative motion-primitive sequences that defer turning until leaving rough terrain, at the cost of increased path length. The sampling-based planner refines trajectory geometry within a fixed route topology, achieving curvature reductions with minimal path extension but exhibiting higher variability across scenarios.

This work contributes a practical framework for embedding wear-conscious objectives into rover motion planning, demonstrating that terrain-aware steering penalties can meaningfully influence trajectory generation without requiring high-fidelity terramechanics models.