ManiSkill 3 beta version releases soon. We strongly recommend new users to wait until that releases. To stay tuned you can star/watch this repository and/or join our discord and/or follow our twitter https://twitter.com/HaoSuLabUCSD. For now you can read about the features below.
Users looking for the original ManiSkill2 can find the commit for that codebase at https://github.com/haosulab/ManiSkill/tree/v0.5.3
ManiSkill is a powerful unified framework for robot simulation and training powered by SAPIEN. The entire stack is as open-source as possible. Among its features, it includes
- GPU parallelized visual data collection system. A policy can collect RGBD + Segmentation data at about 10,000+ FPS with a 4090 GPU, 10-100x faster compared to simulators like Isaac Sim.
- Example tasks covering a wide range of different robot embodiments (quadruped, mobile manipulators, single-arm robots) as well as a wide range of different tasks (table-top, locomotion, scene-level manipulation)
- GPU parallelized tasks, enabling incredibly fast synthetic data collection in simulation at the same or faster speed as other GPU sims like IsaacSim
- GPU parallelized tasks support simulating diverse scenes where every parallel environment has a completely different scene/set of objects
- Flexible task building API that abstracts away much of the complex GPU memory management code
ManiSkill plans to enable all kinds of workflows, including but not limited to 2D/3D vision-based reinforcement learning, imitation learning, sense-plan-act, etc. At the moment the repo contains just state/visual RL baselines, others are coming soon.