Reduction of the Energy Loss in Ground Collisions by Attaching Wobbling Mass to Bipedal Robot

- Bayesian Optimization of Bipedal Robots-

Abstract

Passive dynamic walking provides one of the most efficient gaits of legged robots since the required energy is supplied solely by the gravity. The challenge is to reduce the energy of collision with the ground in order to improve this type of walking to be able to walk on rough ground, e.g., in disaster areas. In this study, we attached a 2DOF wobbling mass to a biped robot and investigated how much the energy consumption due to the ground collision can be reduced. For this purpose, we optimized parameters of the wobbling mass by using Bayesian optimization.