A Control Strategy for Stable Passive Running.pdfVIP

A Control Strategy for Stable Passive Running M. Ahmadi and M. Buehler Centre for Intelligent Machines, Department of Mechanical Engineering McGill University Montreal, Quebec, H3A 2A7 Abstract inclines. Such unforced motion of a mechanical system W epresent a control strategy for a samplified model is called its “passive dynamics.” From the very begin- of a one legged running robot which features compliant ning have Raibert’s robots [14] exploited this principle elements an series with hip and leg actuators. Proper for the vertical motion which is largely produced by a spring selection and initial conditions result in ‘$as- spring-mass system formed from body and compliant sive dynamic” operation close to the desired motion, leg. Others [la, 41 and more recently [B] have suc- without any actuataon. However, this motion is not ceeded in building electrically actuated robots using a stable. The proposed controller is based on online cal- similar design for the vertical dynamics. culations of the desired passive dynamic motion and Only recently has the robotics community inves- stabilises any fixed robot speed. It also tracks large tigated compliant elements to facilitate leg swinging changes in desired robot velocity and remains largely motion. McGeer [7] studied a passive dynamic run- passive for a wide range of velocities, despite a fixed ning biped without torso where two legs were con- set of springs, masses and inertias. To this end, the nected via a spring. Based on linearized numeri- desired motion is expressed as a function of a nor- cal analysis, he proposed strategies to achieve sta- ma1ized “locomotion time” param et er. Comparisons ble steady speed running. Raibert investigated a hip of simulated runs with direct hip actuation show