Shooting-based MPC for Biped Humanoid Robot to Imitate Human Dances

Abstract

In this study, we propose a method for imitating dance movements in biped humanoid robots utilizing a Model Predictive Control (MPC) framework. By employing the MuJoCo MPC framework, we design and implement an MPC controller that serves as a dynamic filter by predefining the cost function. This filter can transform physically unfeasible whole-body motions into feasible movements within the physical constraints of the system. Experimental results demonstrate that this controller can be applied to dance motion capture data, successfully generating feasible dance motions.