Data-Driven Physics Embedded Dynamics with Predictive Control and Reinforcement Learning for Quadrupeds

2026-03-15

AI Summary

This work addresses the challenges of error accumulation and poor interpretability in long-horizon locomotion of quadrupedal robots, which often arise from the absence of physically informed inductive biases in learned dynamics models. To this end, the authors propose embedding a Lagrangian neural network within a reinforcement learning and model predictive control (MPC) framework to construct a dynamics model that adheres to physical principles. They further design an inverse-dynamics-based infinite-horizon MPC formulation to enhance computational efficiency. The resulting approach achieves a favorable balance among dynamical consistency, real-time planning performance, and task execution. Experimental validation on the Unitree Go1 platform demonstrates that, compared to unstructured neural dynamics models, the proposed method yields higher sample efficiency, significantly reduced long-horizon tracking errors, and a fourfold improvement in real-time planning speed.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

State of the art quadrupedal locomotion approaches integrate Model Predictive Control (MPC) with Reinforcement Learning (RL), enabling complex motion capabilities with planning and terrain adaptive behaviors. However, they often face compounding errors over long horizons and have limited interpretability due to the absence of physical inductive biases. We address these issues by integrating Lagrangian Neural Networks (LNNs) into an RL MPC framework, enabling physically consistent dynamics learning. At deployment, our inverse dynamics infinite horizon MPC scheme avoids costly matrix inversions, improving computational efficiency by up to 4x with minimal loss of task performance. We validate our framework through multiple ablations of the proposed LNN and its variants. We show improved sample efficiency, reduced long-horizon error, and faster real time planning compared to unstructured neural dynamics. Lastly, we also test our framework on the Unitree Go1 robot to show real world viability.

Problem

• quadrupedal locomotion
• model predictive control
• reinforcement learning
• physical inductive biases
• long-horizon error

Innovation

• Lagrangian Neural Networks
• Model Predictive Control
• Reinforcement Learning
• Physics-Informed Dynamics
• Quadrupedal Locomotion