Simulation has become a mainstream for both robotics learning and testing the algorithms before applying to the real robots. Recently, NVIDIA’s Omniverse, Isaac Sim powered by Omniverse, and Isaac Gym built on top of Issac Sim have drawn my eyes for robotics learning and testing for deformable linear object manipulation (so far according to NVIDIA Isaac Gym is not integrated into Isaac Sim but rather as a seperated environment, but this will be done soon).
For Issac Gym, NVIDIA has uploaded a technical report paper named Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning on Arxiv for a brief introduction. Here are links for Youtube videos from NVIDIA on Issac Gym:
- Isaac Gym Part 1: Introduction and Getting Started
- Isaac Gym Part 2: Environments, Training, and Tips
- Isaac Gym Part 3A: Academic Labs - University of Toronto
- Isaac Gym Part 3B: Academic Labs - IMLab
- Isaac Gym Part 3C: Academic Labs - Stanford University
- Isaac Gym Part 3D: Academic Labs - Soft-Body Simulation
- Isaac Gym Part 3E: Academic Labs - Eth Zurich
- Isaac Gym Part 4: New Frontiers in End-to-End GPU Accelerated Reinforcement Learning