Optimizing the Shape of Planetary Rover Wheels using the Discrete Element Method and Bayesian Optimization

The Martian Moons eXploration mission will include a rover to conduct in situ science and exploration on the Martian moon Phobos. The rover's locomotion system and especially its wheels require special consideration, as wheeled locomotion has never been employed in a comparably low gravity environment. This paper shows an automated approach to design the shape of rover wheels using high-fidelity simulation models based on the Discrete Element Method and a Bayesian Optimization algorithm. Four main parameters were selected to describe the wheel shape: grouser height, grouser number, grouser chevron angle and rim curvature. On Phobos, three main scenarios are important to ensure the rover's successful operation: Driving forward, backward, and up slopes. The optimization loop thus generates wheels within given limits for the four wheel shape parameters, runs a simulation of a single wheel experiment with free slip conditions for each scenario, and assesses the wheel's performance based on the distance traveled. The found wheel shape performs 64 % better than the previous prototype and, in combination with the found parameter relations, will guide the design of the wheels for the MMX mission