Learning to Walk from Three Minutes of Real-World Data with Semi-structured Dynamics Models

TL;DR

This paper introduces a semi-structured dynamics modeling framework that combines physics-based principles with neural networks, enabling efficient real-world learning of contact-rich systems like quadruped robots with minimal data.

Contribution

The authors develop a semi-structured modeling approach that integrates physics with neural networks, significantly reducing data requirements for contact-rich system dynamics learning.

Findings

Accurate long-horizon predictions with less data.

Successful real-world gait learning on a quadruped robot.

Effective learning on both hard and soft surfaces with minimal data.

Abstract

Traditionally, model-based reinforcement learning (MBRL) methods exploit neural networks as flexible function approximators to represent $\textit{a priori}$ unknown environment dynamics. However, training data are typically scarce in practice, and these black-box models often fail to generalize. Modeling architectures that leverage known physics can substantially reduce the complexity of system-identification, but break down in the face of complex phenomena such as contact. We introduce a novel framework for learning semi-structured dynamics models for contact-rich systems which seamlessly integrates structured first principles modeling techniques with black-box auto-regressive models. Specifically, we develop an ensemble of probabilistic models to estimate external forces, conditioned on historical observations and actions, and integrate these predictions using known Lagrangian dynamics. With this semi-structured approach, we can make accurate long-horizon predictions with substantially less data than prior methods. We leverage this capability and propose Semi-Structured Reinforcement Learning ($\texttt{SSRL}$) a simple model-based learning framework which pushes the sample complexity boundary for real-world learning. We validate our approach on a real-world Unitree Go1 quadruped robot, learning dynamic gaits -- from scratch -- on both hard and soft surfaces with just a few minutes of real-world data. Video and code are available at: https://sites.google.com/utexas.edu/ssrl