Gait Regulation Control Techniques for Robust Legged Locomotion

Abstract

This thesis develops methods of control that allow a multi-legged robot to vary its stepping pattern, the gait of a robot, during locomotion. By constructing feedback control around the gaits a robot may use, we produce behaviors allowing a robot to switch amongst or return to certain gaits while performing feedback control during locomotion. Gait regulation is one specific aspect of gait-based control, and pertains to the use of a control system to monitor and regulate the desired gaits a robot may use. While some gait-based control laws may force a robot to deviate from a nominal gait, gait regulation seeks to return to--or switch amongst--desired gaits as required. After discussing the necessary topological effects of gait regulation control, as well as noting specific constraints that are unique to legged systems, this thesis proposes methods of gait regulation control that place attractors and repellors on a high-dimensional toroidal space, a space relevant to gait timings, in order to converge upon desired gaits. The primary contribution of this thesis is an efficient algorithmic approach to gait regulation that avoids dangerous leg timings while converging to desired gaits, as specified. The system actively manages the basins of convergence for various controllers to achieve a global vector policy directing a robot to certain desired gaits. This work is particularly applicable to four- and six-legged robots, on which a variety of interesting and useful gait timings exist. Specifically, we apply gait regulation to a climbing hexapod, on which we design a climbing behavior based upon a collection of reactive force control techniques, causing the robot to deviate from its desired gait. With gait regulation, the robot maintains use of its desired gaits, with the additional ability to actively transition amongst gaits while climbing.