Whole-Body Model-Predictive Control of Legged Robots with MuJoCo

2025-03-06

AI Summary

This work addresses the challenge of deploying real-time whole-body model predictive control (MPC) for legged robots on physical hardware. We propose an end-to-end solution that eliminates the need for sim-to-real adaptation. Methodologically, we develop a lightweight iterative linear quadratic regulator (iLQR)-based real-time optimization framework, leveraging MuJoCo for dynamics modeling and finite-difference approximations for Jacobians. All controllers are trained exclusively in simulation. To our knowledge, this is the first demonstration of direct transfer of purely simulation-trained whole-body MPC to physical quadrupedal, bipedal, and full-scale humanoid robots. Experiments confirm real-time performance (≥100 Hz) and robustness across dynamic walking and stable gait tasks. The open-sourced code and hardware demonstration videos significantly lower the barrier to deploying whole-body MPC on real legged platforms.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

We demonstrate the surprising real-world effectiveness of a very simple approach to whole-body model-predictive control (MPC) of quadruped and humanoid robots: the iterative LQR (iLQR) algorithm with MuJoCo dynamics and finite-difference approximated derivatives. Building upon the previous success of model-based behavior synthesis and control of locomotion and manipulation tasks with MuJoCo in simulation, we show that these policies can easily generalize to the real world with few sim-to-real considerations. Our baseline method achieves real-time whole-body MPC on a variety of hardware experiments, including dynamic quadruped locomotion, quadruped walking on two legs, and full-sized humanoid bipedal locomotion. We hope this easy-to-reproduce hardware baseline lowers the barrier to entry for real-world whole-body MPC research and contributes to accelerating research velocity in the community. Our code and experiment videos will be available online at:https://johnzhang3.github.io/mujoco_ilqr

Problem

• Real-time whole-body MPC for legged robots
• Generalization of simulation policies to real-world
• Lowering entry barrier for real-world MPC research

Innovation

• iLQR algorithm with MuJoCo dynamics
• Real-time whole-body MPC on hardware
• Sim-to-real generalization with minimal adjustments