Exploiting Passive Stability for Hierarchical Control

The dynamics of a Spring Loaded Inverted Pendulum (SLIP) \template” [1] approximate well the center of mass (COM) of running animals, humans, and of the robot RHex [2]. Running control can therefore be ierarchically structured as a high level SLIP control and the anchoring of SLIP in the complex morphology of the physical system. Analysis of the sagittal plane lossless SLIP model has shown that it includes parameter regions where its gait is passively stabilized, i.e. with the discrete control input | the leg touchdown angle | held constant. We present numerical evidence to suggest that an open loop \clock” excitation of a high degree of freedom hexapedal robot model can lead to asymptotically stable limit cycles that \anchor” [1] the SLIP model in its self stabilizing regime. This motivates the search for completely feedforward SLIP locomotion control strategies, which we now speculate may be successfully used to elicit a self-stabilizing running robot such as RHex. For more information: Kod*Lab Comments BibTeX entry @inproceedings{altendorfer-CLAWAR-2002, author = {R. Altendorfer and R. M. Ghigliazza and P. Holmes and D. E. Koditschek}, title = {Exploiting passive stability for hierarchical control}, booktitle = {Proceedings of the Fifth International Conference on Climbing and Walking Robots, Paris, France}, year = {2002} } This conference paper is available at ScholarlyCommons: http://repository.upenn.edu/ese_papers/667 Exploiting passive stability for hierarchical control R Altendorfer Artificial Intelligence Lab, University of Michigan, Ann Arbor, MI 48109, US R M Ghigliazza Dept. of Mech. and Aerospace Eng., Princeton University, Princeton, NJ 08544, US P Holmes Dept. of Mech. and Aerospace Eng., Princeton University, Princeton, NJ 08544, US Program in Applied and Computational Mathematics, Princeton University, Princeton, NJ 08544, US D E Koditschek Artificial Intelligence Lab, University of Michigan, Ann Arbor, MI 48109, US ABSTRACT The dynamics of a Spring Loaded Inverted Pendulum (SLIP) “template” [1] approximate well the center of mass (COM) of running animals, humans, and of the robot RHex [2]. Running control can therefore be hierarchically structured as a high level SLIP control and the anchoring of SLIP in the complex morphology of the physical system. Analysis of the sagittal plane lossless SLIP model has shown that it includes parameter regions where its gait is passively stabilized, i.e. with the discrete control input — the leg touchdown angle — held constant. We present numerical evidence to suggest that an open loop “clock” excitation of a high degree of freedom hexapedal robot model can lead to asymptotically stable limit cycles that “anchor” [1] the SLIP model in its self stabilizing regime. This motivates the search for completely feedforward SLIP locomotion control strategies, which we now speculate may be successfully used to elicit a self-stabilizing running robot such as RHex.The dynamics of a Spring Loaded Inverted Pendulum (SLIP) “template” [1] approximate well the center of mass (COM) of running animals, humans, and of the robot RHex [2]. Running control can therefore be hierarchically structured as a high level SLIP control and the anchoring of SLIP in the complex morphology of the physical system. Analysis of the sagittal plane lossless SLIP model has shown that it includes parameter regions where its gait is passively stabilized, i.e. with the discrete control input — the leg touchdown angle — held constant. We present numerical evidence to suggest that an open loop “clock” excitation of a high degree of freedom hexapedal robot model can lead to asymptotically stable limit cycles that “anchor” [1] the SLIP model in its self stabilizing regime. This motivates the search for completely feedforward SLIP locomotion control strategies, which we now speculate may be successfully used to elicit a self-stabilizing running robot such as RHex.