Graph Networks as Learnable Physics Engines for Inference and Control
Alvaro Sanchez-Gonzalez Nicolas Heess Jost Tobias Springenberg Josh Merel Martin Riedmiller Raia Hadsell Peter Battaglia Alvaro Sanchez-Gonzalez Nicolas Heess Jost Tobias Springenberg Josh Merel Martin Riedmiller Raia Hadsell Peter Battaglia
Abstract
Understanding and interacting with everyday physical scenes requires rich knowledge about the structure of the world, represented either implicitly in a value or policy function, or explicitly in a transition model. Here we introduce a new class of learnable models—based on graph networks—which implement an inductive bias for object- and relation-centric representations of complex, dynamical systems. Our results show that as a forward model, our approach supports accurate predictions from real and simulated data, and surprisingly strong and efficient generalization, across eight distinct physical systems which we varied parametrically and structurally. We also found that our inference model can perform system identification. Our models are also differentiable, and support online planning via gradient-based trajectory optimization, as well as offline policy optimization. Our framework offers new opportunities for harnessing and exploiting rich knowledge about the world, and takes a key step toward building machines with more human-like representations of the world.
中文速览
用图神经网络（Graph Network，GN）来学习复杂物理系统的动力学，一直是个棘手的问题——既要泛化到没见过的结构，又要精确预测真实的物理行为。研究者提出了一套基于图网络的可学习框架：把物理系统里的物体映射为图的节点、关节映射为边，让网络在节点和边上共享参数，从而天然地具备「以物体和关系为中心」的归纳偏置（inductive bias）。在七个MuJoCo仿真环境和一个真实机械臂场景中，该前向模型不仅能准确预测下一时刻状态，还能在从未见过的系统参数和拓扑结构（如不同节数的游泳体）上实现强泛化；配套的推理模型可以从观测轨迹中隐式识别系统参数，控制算法则借助模型的可微性实现基于梯度的轨迹优化和强化学习，性能与强基线相当甚至更优。这项工作意义在于，它是首个无需硬编码物理定律、却能处理复杂三维物理系统的通用可学习物理引擎，为构建具有人类式世界模型的智能体迈出了关键一步。
原文 arXiv:1806.01242;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1806.01242v1