Deep Reinforcement Learning of Multi-Modal Locomotion for Wheel-on-Limb Robotics

As the planetary science community becomes increasingly interested in the exploration of extreme terrain environments like lava tubes and icy moon surfaces, the need for novel rover mobility systems becomes increasingly apparent. Wheel-on-limb systems have proven themselves to be flexible and capable of efficient motion and navigation through rough terrain, but the autonomous capabilities of such systems are presently limited. As such robots are necessarily high degree-of-freedom systems, major challenges exist in the switching between different locomotion modalities and in motion planning while exploring due to the large rover configuration and planning spaces. This project proposes the development of an online motion planning system for enabling autonomous multi-modal locomotion of wheel-on-limb systems over rough terrain. A kinodynamic model of a wheel-on-limb robot will first be compiled in a design configuration relevant to NASA’s current exploration goals, then a mockup created in simulation to be commanded by software written in the Robot Operating System. Model-based controllers for the primary modes of locomotion—actively articulated suspension, wheel-stepping, and more exotic gaits like “inchworming”—will be written and coordinated using a novel gait-switching controller to enable smooth transition between locomotion modalities. Lastly, a motion planner will be trained using deep reinforcement learning to enable the robot while perceiving and exploring its environment using depth cameras to perceive a procedurally generated elevation map. If an existing wheel-on-limb system is available for limited study through the paired NASA center, limited tests on hardware can be applied to create more accurate simulations through dynamics analysis and/or demonstrate the efficacy of the motion planner and controller architecture in limited field trials. Over the course of this four-year project, end-to-end autonomy for the wheel-on-limb mobility system will be achieved and will build upon the capabilities of architecture of interest to NASA.

This is early/mid-stage (TRL 2) — the most realistic path in is NASA SBIR/STTR, which funds small businesses and research institutions to develop technology aligned with NASA's needs (equity-free, phased funding). Check whether a current SBIR/STTR solicitation topic overlaps with this project's technology area, or contact the project directly (above) to ask.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.