End-to-End Differentiable Physics for Learning and Control

TLDR

The paper introduces a differentiable physics engine that can be embedded in deep neural networks for end‑to‑end learning. The engine performs analytic backpropagation through a linear complementarity problem simulator, avoiding finite‑difference gradients and enabling greater flexibility. Experiments demonstrate that the engine can learn physical parameters, accurately simulate observed behavior, and support gradient‑based control, achieving high sample efficiency; code is provided.

Abstract

We present a differentiable physics engine that can be integrated as a module in deep neural networks for end-to-end learning. As a result, structured physics knowledge can be embedded into larger systems, allowing them, for example, to match observations by performing precise simulations, while achieves high sample efficiency. Specifically, in this paper we demonstrate how to perform backpropagation analytically through a physical simulator defined via a linear complementarity problem. Unlike traditional finite difference methods, such gradients can be computed analytically, which allows for greater flexibility of the engine. Through experiments in diverse domains, we highlight the system's ability to learn physical parameters from data, efficiently match and simulate observed visual behavior, and readily enable control via gradient-based planning methods. Code for the engine and experiments is included with the paper.