Isaac Gym - Preview Release
NVIDIA’s physics simulation environment for reinforcement learning research.
End-to-End GPU accelerated
Physics simulation in Isaac Gym runs on the GPU, storing results in PyTorch GPU tensors
Thousand of Environments
Using the Isaac Gym tensor-based APIs, observations and rewards can be calculated on the GPU in PyTorch, enabling thousands of environments to run in parallel on a single workstation
Isaac Gym features include:
- Support for importing URDF and MJCF files with automatic convex decomposition of imported 3D meshes for physical simulation
- GPU accelerated tensor API for evaluating environment state and applying actions
- Support for a variety of environment sensors - position, velocity, force, torque, etc
- Runtime domain randomization of physics parameters
- Jacobian / inverse kinematics support
An initial release of tensor-based Gym APIs for GPU accelerated RL is now available as part of the NVIDIA Omniverse Isaac Sim 2022.1 robotics simulator. Work is ongoing to continue improving Omniverse Isaac Gym RL functionality.
Until Omniverse Isaac Gym functionality is feature complete, this standalone Isaac Gym Preview release will remain available to facilitate the work of researchers and academics who want to explore the potential of GPU-based reinforcement learning.
A variety of examples and GPU accelerated training environments are also available:
Isaac Gym Preview 4 Release:
- This release aligns the PhysX implementation in standalone Preview Isaac Gym with Omniverse Isaac Sim 2022.1 to simplify migration to Omniverse for RL workloads
- Added support for SDF collisions with a nut & bolt example
- Additional Factory RL samples available in the https://github.com/NVIDIA-Omniverse/IsaacGymEnvs repository.
- Enabled gyroscopic forces by default to improve simulation
- Allowing customizing rest_offset and contact_offset per asset and per individual shape.
- Added parsing of spherical (ball) joints support in URDF importer
- Various bug fixes for elastic collision behaviour, resetting fixed base actors, friction randomization, and friction mode settings
Note that limited support will be available for this preview prior to the release of tensor-based Gym API support in Omniverse.
Ready to get started with Isaac Gym?Join now