Meeting Abstract
We use a simple walking robot, referred to as a Dynamic Control Platform (DCP) to investigate algorithms for the control of bipedal walking. Our control strategies target orthogonal constraint — a perpendicular relationship between the center of mass velocity vector and ground reaction force vector. This is done by actuating the ‘ankle’ joint to achieve braking or propulsion and thereby modulate the direction of the ground reaction force vector. The ‘ankles’ are driven by DC gear motors through a belt-pulley system and the symmetrical heel and toe extensions of the foot are scaled to the length of a human foot. A force-instrumented trackway provides inputs via Wi-Fi to the control system on-board the robot. Deviation from orthogonal constraint is determined each millisecond in by 1) calculating the ground reaction force vector and 2) integrating the resulting acceleration to determine the velocity vector. We use mechanical cost analysis to analyze the walking dynamics of the DCP in comparison to human walking dynamics. This analysis determines the mechanical cost of transport (CoTmech) of a walking stride, as well individual instances of high and low cost throughout the single and double stance phases of walking. While humans show appreciable mechanical cost in both phases of the stride, comparable measurements have not yet been made on a walking robot. We explore constraints and solutions with respect to this cost profile to inform the mechanism and control of bipedal walking. Control algorithms developed in the DPC can reveal new strategies for bipedal walking gaits in robots, interpret effects of foot length, compliance or other mechanical properties, and improve the function of powered prosthetics or exoskeletons.
