Trajectory Generation of Center of Mass for Biped Robot by Monte Carlo Model Predictive Control Considering Upper Body Contact with Walls and Constraints of ZMP

Abstract

Biped robots have a body structure similar to that of humans, and therefore have a high affinity for environments and tasks set for humans. Therefore, they are expected to be utilized in particularly hazardous disaster sites. In the process of accomplishing a walking task at a disaster site, contact between the body and the outside world may be unavoidable, and there are cases where conventional online trajectory generation of center of mass without prior consideration of contact between the body and the outside world cannot handle such a situation. In this paper, we extend the conventional method to the control considering contact with the wall surfaces by using MCMPC, a sample-based method of model predctive control, and confirm that it shows better performance than the control without considering contact with the wall surfaces. Furthermore, we confirmed that the proposed method can be executed in real time under the conditions we set.