This project explores new locomotion techniques that will enable robotic systems to recover from a variety of failures. The primary focus is on a class of robot locomotion termed dynamically coupled locomotion. The most familiar forms of locomotion, such as human walking or automobile rolling, employ kinematic coupling. But locomotory systems are subject to error conditions such as falling or becoming high-centered. Recovery from such errors sometimes requires locomotion modes that, while less efficient, are nonetheless essential to the robustness and overall competence of the robotic system. One example would be a six-legged robot that has become high-centered on an obstacle such that its legs no longer reach the ground -- by aggressively moving its legs it is possible to excite vibration modes that cause the robot to gradually move off of the obstacle. The specific goal of this project is to explore the theoretical and practical capabilities of such locomotion modes, to survey their potential application in mobile robotic systems, and to demonstrate their usefulness in real robots. The main theoretical thrust of the work is to characterize the motion capabilities of dynamically coupled locomotion, and to synthesize motion patterns for mobile robots to recover from errors. Experimental and simulation work will serve to demonstrate and validate the research. Surveys of mobile robot structures and failure modes will provide context and grounding to assess the potential usefulness of the work.
