Information for the paper accepted for LivingMachines2020

(Springer Link) [https://doi.org/10.1007/978-3-030-64313-3_21]

Split-belt Adaptation Model of a Decerebrate Cat Using a Quadruped Robot with Learning

K. Kodono and H. Kimura

Abstract

We propose a model connecting body dynamics and sensor feedback to investigate gait adaptation mechanisms of decerebrate cats in split-belt waking. In our previous studies, we proposed a leg controller using leg loading/unloading for the leg phase transition of a quadruped robot:``Kotetsu.'' The purpose of this study is to make a model of split-belt gait adaptation in spinal cats and decerebrate cats referring to biological knowledge and to evaluate the validity of those models using Kotetsu. We construct the spinal cat model integrating our leg controller with Frigon's spinal cord model. In addition, we employ motor learning in the cerebellum by long-term depression for the decerebrate cat model. As the results of experiments, we show that early adaptation in split-belt walking is obtained by the stance-to-swing leg phase transition mechanism in the spinal cat model. We also show that late adaptation in split-belt walking is obtained motor learning as the step distance adjustment in the decerebrate cat model. The validity of those models are evaluated by comparing durations of forelegs bisupport phases and duty ratios in the transition from tied-belt to split-belt walking between decerebrate cats and Kotetsu with the decerebrate cat model.


Constructing the model of the gait adaptation mechanism of decerebrate cats in split-belt walking using a quadruped robot


Spec. of Kotetsu
History and abstract of the study using Kotetsu & Kurama