ORBIT: A Unified Simulation Framework for Interactive Robot Learning Environments

  • 2023-01-10 20:19:17
  • Mayank Mittal, Calvin Yu, Qinxi Yu, Jingzhou Liu, Nikita Rudin, David Hoeller, Jia Lin Yuan, Pooria Poorsarvi Tehrani, Ritvik Singh, Yunrong Guo, Hammad Mazhar, Ajay Mandlekar, Buck Babich, Gavriel State, Marco Hutter, Animesh Garg
  • 47

Abstract

We present ORBIT, a unified and modular framework for robot learning poweredby NVIDIA Isaac Sim. It offers a modular design to easily and efficientlycreate robotic environments with photo-realistic scenes and fast and accuraterigid and deformable body simulation. With ORBIT, we provide a suite ofbenchmark tasks of varying difficulty -- from single-stage cabinet opening andcloth folding to multi-stage tasks such as room reorganization. To supportworking with diverse observations and action spaces, we include fixed-arm andmobile manipulators with different physically-based sensors and motiongenerators. ORBIT allows training reinforcement learning policies andcollecting large demonstration datasets from hand-crafted or expert solutionsin a matter of minutes by leveraging GPU-based parallelization. In summary, weoffer an open-sourced framework that readily comes with 16 robotic platforms, 4sensor modalities, 10 motion generators, more than 20 benchmark tasks, andwrappers to 4 learning libraries. With this framework, we aim to supportvarious research areas, including representation learning, reinforcementlearning, imitation learning, and task and motion planning. We hope it helpsestablish interdisciplinary collaborations in these communities, and itsmodularity makes it easily extensible for more tasks and applications in thefuture. For videos, documentation, and code: https://isaac-orbit.github.io/.

 

Quick Read (beta)

loading the full paper ...