ManiSkill

ManiSkill

ManiSkill is an open-source GPU-parallel robotic simulation and training framework.

Agent SkillDevOpen source
Type
Agent Skill
Open source
Yes
GitHub Stars
★ 3.0k
Source
skill-github

Overview

ManiSkill is an open-source framework based on SAPIEN, focused on robotic manipulation skills. It supports GPU-parallel visual data collection, GPU-parallel simulation, and heterogeneous simulation. The framework provides various task examples involving different types of robots and tasks, along with a flexible task construction API. ManiSkill also enables policy deployment from simulation to the real world. Users can install ManiSkill via pip and set up Vulkan for rendering. It is suitable for researchers and developers needing robotic simulation and training.

Capabilities

  • GPU-parallel visual data collection
  • GPU-parallel simulation
  • Heterogeneous simulation
  • Multiple robot and task examples
  • Flexible task construction API
  • Policy deployment from simulation to the real world

Use cases

Robotic simulation and trainingVisual data collectionSynthetic data generationPolicy evaluation from simulation to the real world

Setup

Requires: PythontorchVulkan
pip install --upgrade mani_skill && pip install torch && set up Vulkan rendering environment

This information was compiled by AI from public sources and may contain inaccuracies — please refer to the source.

FAQ

What systems does ManiSkill support?

Currently primarily supports Linux and NVIDIA GPUs; Windows and MacOS support is limited.

How do I get started with ManiSkill?

Please refer to the quick start documentation or Colab notebooks.

Related skills