Robust Dynamic Locomotion of Heavy Hexapod Robots via NMPC and Weighted WBC

Large-scale, heavy-payload hexapod robots show great potential for industrial applications. However, when subjected to external force disturbances, they find it difficult to maintain stable locomotion due to their bulky structure and high inertia. To address this challenge, we propose a model-based control framework that combines a contact-aware centroidal nonlinear model predictive controller (NMPC) with a weighted inverse-dynamics whole-body controller (WBC). The NMPC serves as a centroidal planner that accounts for passive, ball-jointed large footplates and generates contact-aware centroidal trajectories and ground-reaction forces in real time, while the WBC solves a quadratic program to track these references under actuator torque and friction-polytope constraints. High-fidelity MuJoCo simulations were conducted on a floating-base hexapod carrying an additional 5 t payload (total mass 10 t) and subjected to external push perturbations, demonstrating that the controller limits the peak CoM deviation to about 0.05 m and restores nominal locomotion within 1.96 s.