paper-with-me

홈 › Papers

Graph networks as learnable physics engines for inference and control

2018-06-04 · ICML 2018 7 · Alvaro Sanchez-Gonzalez, Nicolas Heess, Jost Tobias Springenberg, Josh Merel, Martin Riedmiller, Raia Hadsell, Peter Battaglia

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.

📄 PDF Abstract BibTeX arXiv:1806.01242

Code (1)

fxia22/gn.pytorch pytorch

Tasks

Inductive BiasWeather Forecasting

Similar Papers 제목 키워드 기반

Propagation Networks for Model-Based Control Under Partial Observation

2018-09-28 · Yunzhu Li, Jiajun Wu, Jun-Yan Zhu, Joshua B. Tenenbaum 외

There has been an increasing interest in learning dynamics simulators for model-based control. Compared with off-the-shelf physics engines, a learnable simulator can quickly adapt to unseen objects, scenes, and tasks. Ho…

Deep Reinforcement LearningReinforcement Learning

Can Reinforcement Learning for Continuous Control Generalize Across Physics Engines?

2020-10-27 · Aaqib Parvez Mohammed, Matias Valdenegro-Toro

Reinforcement learning (RL) algorithms should learn as much as possible about the environment but not the properties of the physics engines that generate the environment. There are multiple algorithms that solve the task…

continuous-controlContinuous ControlDeep Reinforcement LearningMuJoCo+3

Efficient Probabilistic Inference in the Quest for Physics Beyond the Standard Model

2018-07-20 · NeurIPS 2019 12 · Atılım Güneş Baydin, Lukas Heinrich, Wahid Bhimji, Lei Shao 외

We present a novel probabilistic programming framework that couples directly to existing large-scale simulators through a cross-platform probabilistic execution protocol, which allows general-purpose inference engines to…

Bayesian InferenceProbabilistic Programming

Physics-as-Inverse-Graphics: Unsupervised Physical Parameter Estimation from Video

2019-05-27 · ICLR 2020 1 · Miguel Jaques, Michael Burke, Timothy Hospedales

We propose a model that is able to perform unsupervised physical parameter estimation of systems from video, where the differential equations governing the scene dynamics are known, but labeled states or objects are not …

Inductive BiasModel Predictive Controlparameter estimationScene Understanding+1

Combining Physical Simulators and Object-Based Networks for Control

2019-04-13 · Anurag Ajay, Maria Bauza, Jiajun Wu, Nima Fazeli 외

Physics engines play an important role in robot planning and control; however, many real-world control problems involve complex contact dynamics that cannot be characterized analytically. Most physics engines therefore e…

Object