Developments in Nonlinear Model Predictive Control for Telescopic-Wheeled-Legged Robot Tachyon 3

Abstract

This work presents advancements in nonlinear model predictive control (NMPC) for the telescopic-wheeled-legged robot, Tachyon 3. Our NMPC is based on the full-centroidal NMPC formulation to accurately capture the complex constraints of Tachyon 3. Furthermore, we introduce a stochastic NMPC to ensure safety-related constraints even in the presence of contact uncertainties. The proposed NMPC is implemented with a real-time CBF-QP controller to ensure strict safety and an internal state integrator to adapt the NMPC with actuators that employ high-gain position control. The effectiveness of our NMPC is demonstrated through hardware experiments with limited on-board computation.