A quadruped robot with four actuated hip joints and four passive highly compliant knee joints is used to demonstrate the potential of underactuation from two standpoints： learning locomotion and perception. First, we...A quadruped robot with four actuated hip joints and four passive highly compliant knee joints is used to demonstrate the potential of underactuation from two standpoints： learning locomotion and perception. First, we show that： （i） forward locomotion on flat ground can be learned rapidly （minutes of optimization time）; （ii） a simulation study reveals that a passive knee configuration leads to faster, more stable, and more efficient locomotion than a variant of the robot with active knees; （iii） the robot is capable of learning turning gaits as well. The merits of underactuation （reduced controller complexity, weight, and energy consumption） are thus preserved without compromising the versatility of behavior. Direct optimization on the reduced space of active joints leads to effective learning of model-free controllers. Second, we find passive compliant joints with po- tentiometers to effectively complement inertial sensors in a velocity estimation task and to outperform inertial and pressure sensors in a terrain detection task. Encoders on passive compliant joints thus constitute a cheap and compact but powerful sensing device that gauges joint position and force/torque, and -- if mounted more distally than the last actuated joints in a legged robot -- it delivers valuable information about the interaction of the robot with the ground.展开更多
基金Acknowledgment Matej Hoffmann was supported by the Swiss National Science Foundation project ＂From locomotion to cognition＂ （Grant No. 200020-122279/1）. Jakub Simanek was supported by the Grant Agency of the CTU in Prague （Grant No. SGS 15/163/OHK3/2T/13）. Matej Hoffmann would like to thank Roll Pfeifer for continuous support of this project and to the collaborators that contributed to the investigations that laid the foundations for this work, in particular Fumiya Iida, Michal Reinstein, Nico Schmidt, and students Stefan Hutter, Richard Meuris, Nicolas Ruegg, Urs Fassler, and Mathias Weyland. We would also like to thank Koh Hosoda for the idea that passive joints may increase the overall ground contact duration of individual legs and Nadja Schilling for a discussion of the ＂template＂ of leg morphology in mammalian running. Finally, we are indebted to Michal Reinstein and Kenichi Narioka for valuable comments on the manuscript.
文摘A quadruped robot with four actuated hip joints and four passive highly compliant knee joints is used to demonstrate the potential of underactuation from two standpoints： learning locomotion and perception. First, we show that： （i） forward locomotion on flat ground can be learned rapidly （minutes of optimization time）; （ii） a simulation study reveals that a passive knee configuration leads to faster, more stable, and more efficient locomotion than a variant of the robot with active knees; （iii） the robot is capable of learning turning gaits as well. The merits of underactuation （reduced controller complexity, weight, and energy consumption） are thus preserved without compromising the versatility of behavior. Direct optimization on the reduced space of active joints leads to effective learning of model-free controllers. Second, we find passive compliant joints with po- tentiometers to effectively complement inertial sensors in a velocity estimation task and to outperform inertial and pressure sensors in a terrain detection task. Encoders on passive compliant joints thus constitute a cheap and compact but powerful sensing device that gauges joint position and force/torque, and -- if mounted more distally than the last actuated joints in a legged robot -- it delivers valuable information about the interaction of the robot with the ground.
