Toward Human-Robot Collaborative Dancing: Trajectory Optimization for Hybrid Robotic System

Abstract

This paper presents an approach for trajectory optimization of a humanoid robot performing collaborative dancing with a human partner. The challenge lies in generating optimal robot control that can mimic human dance. The work is part of an ongoing Human-Robot Partner Dancing project that aims to establish safe interaction between a robot and a human during salsa dancing. The proposed method utilizes a hybrid iterative Linear Quadratic Regulator (Hybrid-iLQR) that incorporates the saltation matrix to handle the discrete transitions between continuous dynamics modes. This allows optimizing trajectories of high degree-of-freedom, nonlinear hybrid dynamic systems that need to transition between different dynamic modes involving hybrid events like ground contacts. The simulation results demonstrate the ability to generate optimized trajectories that can potentially enable fluid human-robot dance collaboration. The framework and tools generated in this paper serve as a proof of concept that can be applied to real-life bipedal robot in the future.