Quasi-static rolling control for hybrid rolling-walking and climbing robots
Motivated by the need for greater speed, efficiency and adaptability in climbing and walking robots, a class of legged robots have been developed which compliment their walking and climbing capabilities with rolling. Rolling capabilities are provided by innovative morphologies (including circular cross-section exoskeletons), without the need for additional resources beyond those required for walking and climbing. Herein is presented the design of two such robots, the development of quasi-static rolling locomotion controllers for them, and a comparison of experimentally obtained speed and energy data for walking versus rolling locomotion. Rather than basing the control law on simulated model based results as in previous work, the quasi-static control law developed here relies on the passive stability of a disproportionately weighted circular shape. On an even slope, such a shape will come to equilibrium where the center of gravity (CG) is on the line connecting the geometric center of the circle and the ground contact point. Sufficiently slow perturbations to the CG position will cause the circle to roll to restore equilibrium; therefore slowly moving the CG in some polar trajecotry of nonzero length from 0°—>360° around the geometric center of the circular shape will cause it to roll a complete revolution. Joint paths to produce an orbit of the CG about the geometric center of the robot's circular shape are generated using a static optimization routine. The intent of this routine is to find smooth joint angle paths that produce a path of the CG at some specified distance L from the geometric center at angles from 0°—>360°. Joint paths are generated offline for various L values and played back at different rates in order to evaluate the validity of the control. Experiments on the Rolling Disk Biped (RDB) robot show that in general, trajectory tracking of a linearly increasing rolling angle reference is improved as L increases and as the playback rate decreases which is in concert with our quasi-static assumption. This thesis will show that a feed-forward quasi-static rolling control law may be used to successfully increase the energy efficiency of a legged mobile robot on a level surface (by as much as 5.5 times that of walking on the RDB robot). It will also be shown that such a quasistatic control law may be used to reliably track a linearly increasing rolling trajectory for a limited range of rolling rates and slopes on the RDB.
University of Utah;
University of Utah;
Relation-Is Version Of
Digital reproduction of “Quasi-static rolling control for hybrid rolling-walking and climbing robots” J. Willard Marriott Library Special Collections TJ7.5 2008 .P48