Robust Quadrupedal Pronking Through Layered Model Predictive and Template-Based Control

This paper presents a robust control framework for quadrupedal pronking by integrating template-based control with model predictive control (MPC) to achieve stable and adaptable locomotion. We begin by formulating a template model that encapsulates the core dynamics of pronking, generating reference motions to guide the robot’s gait. A novel linear time varying MPC strategy is designed to track these references and compute control inputs, while a nonlinear model predictive controller (NMPC) handles foot placement planning based on the template’s dynamic equations. Our approach is implemented and validated through MuJoCo simulations, where we evaluate its performance against existing methods. Stability and robustness are rigorously assessed through analysis of convergence across varied initial conditions and disturbance rejection tests. The results demonstrate that this method effectively produces stable, robust pronking motions, illustrating the potential of combining template-based and predictive control strategies for advanced quadrupedal locomotion.