Back to Genesis

Genesis World

README.md

1.1.023.0 KB
Original Source

Genesis World

Genesis World is a simulation platform for physical AI developments. It combines a unified multi-physics engine, a photo-realistic renderer (Nyx), and a cross-platform compiler (Quadrants) behind a Pythonic simulation interface. Genesis World is designed to scale from a single laptop kernel to datacenter-grade GPUs, while remaining easy to read, extend, and embed in research code.

It was previously named Genesis and started as an academic project since Dec 2024, and its development is now officially supported by Genesis AI.

For more technical details, refer to our blog post.

Table of Contents

  1. What is Genesis World?
  2. Catalogue
  3. Quick Installation
  4. Docker
  5. Contribution
  6. Support
  7. License and Acknowledgments
  8. Citation

What is Genesis World?

Genesis World occupies the four layers inside the dashed box. Above sits whatever you build (robotics environments, ML pipelines, data generation, agentic simulation); below sits whatever compute backend you have.

  • Simulation Interface — the user-facing API: asset parsing (URDF, MJCF, OBJ, GLB, USD, …), entity accessors, controllers, sensors, parallel and heterogeneous environments, and a built-in GUI.
  • Physics — a unified multi-physics engine integrating Rigid, FEM, MPM, Particle (PBD / SPH), uipc, an explicit coupler, and SAP, all sharing one scene and one state.
  • Render — three rendering paths plug in as camera sensors: Nyx (our in-house renderer designed for robotics), Luisa (DSL ray tracer), and Pyrender (rasterizer).
  • CompilerQuadrants lowers Python kernel code to CUDA, AMD ROCm, Apple Metal, Vulkan, x86, and ARM64. It carries Genesis's autodiff, GPU graphs, and fastcache machinery.

Documentation

Catalogue

Three sections, mirroring the Genesis layers that ship runnable demos: Physics (solvers and multi-solver coupling), Rendering (in-repo camera setups plus the Nyx walkthroughs hosted in genesis-nyx), and Simulation Interface (sensors, GUI, controllers, parallel/heterogeneous envs, and tutorials). Most scripts run end-to-end after pip install -e ".[dev]"; demos that depend on optional backends (e.g. the IPC and Nyx examples) need the extras listed in Optional extras.

Physics

Rigid: franka cubeRigid: collision towerRigid: contype
FEM: hard & soft constraintMPM: tutorialMPM: sand wheel
SPH: rigidSPH: + MPMPBD: liquid
PBD: clothStable Fluid: smokeIPC: robot cloth teleop
Coupler: cloth on rigidCoupler: rigid + MPMCoupler: cut dragon
Coupler: water wheelCoupler: flush cubesSAP: Franka grasp rigid cube

Rendering

Genesis exposes three rendering paths as camera sensors: built-in (Nyx / Luisa / Pyrender) and detailed Nyx walkthroughs hosted in genesis-nyx.

Follow entityAnimated cameraNyx: hello
Nyx: attached cameraNyx: PBR materialsNyx: light types
Nyx: 3D Gaussian splatNyx: object pickingNyx: multi-cam multi-env

Simulation Interface

Controlling a robotGUI: ImGui joint controlHeterogeneous envs
Domain randomizationSensor: depth cameraSensor: IMU
Sensor: lidarSensor: tactile sandboxSensor: contact force
Sensor: surface distanceSensor: temperature gridGUI: debug drawing
GUI: mesh point pickerGUI: mouse interactionDiff-IK controller
Batched IKDroneAdvanced: worm

Quick Installation

Using pip

Install PyTorch first following the official instructions.

Then, install Genesis via PyPI:

bash
pip install genesis-world  # Requires Python>=3.10,<3.14;

For the latest version to date, make sure that pip is up-to-date via pip install --upgrade pip, then run command:

bash
pip install git+https://github.com/Genesis-Embodied-AI/genesis-world.git

Note that the package must still be updated manually to sync with main branch.

Users seeking to contribute are encouraged to install Genesis in editable mode. First, make sure that genesis-world has been uninstalled, then clone the repository and install locally:

bash
git clone https://github.com/Genesis-Embodied-AI/genesis-world.git
cd genesis-world
pip install -e ".[dev]"

It is recommended to systematically execute pip install -e ".[dev]" after moving HEAD to make sure that all dependencies and entrypoints are up-to-date.

Optional extras

IPC solver (uipc backend)pip install pyuipc (Linux / Windows x86, NVIDIA GPU)
Nyx rendererpip install gs-nyx — see genesis-nyx

Quadrants is bundled with Genesis automatically; no extra install. The standalone wheel (pip install quadrants) is documented at Quadrants for users who want the compiler outside Genesis.

Using uv

uv is a fast Python package and project manager.

Install uv:

bash
# On macOS and Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# On Windows
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Quick start with uv:

bash
git clone https://github.com/Genesis-Embodied-AI/genesis-world.git
cd genesis-world
uv sync

Then install PyTorch for your platform:

bash
# NVIDIA GPU (CUDA 12.6 as an example)
uv pip install torch --index-url https://download.pytorch.org/whl/cu126

# CPU only (Linux/Windows)
uv pip install torch --index-url https://download.pytorch.org/whl/cpu

# Apple Silicon (Metal/MPS)
uv pip install torch

Run an example:

bash
uv run examples/rigid/single_franka.py

Docker

If you want to use Genesis from Docker, you can first build the Docker image as:

bash
docker build -t genesis -f docker/Dockerfile docker

Then you can run the examples inside the docker image (mounted to /workspace/examples):

bash
xhost +local:root # Allow the container to access the display

docker run --gpus all --rm -it \
-e DISPLAY=$DISPLAY \
-e LOCAL_USER_ID="$(id -u)" \
-v /dev/dri:/dev/dri \
-v /tmp/.X11-unix/:/tmp/.X11-unix \
-v $(pwd):/workspace \
--name genesis genesis:latest

AMD users

AMD users can use Genesis using the docker/Dockerfile.amdgpu file, which is built by running:

docker build -t genesis-amd -f docker/Dockerfile.amdgpu docker

and can then be used by running:

xhost
docker run -it --network=host \
 --device=/dev/kfd \
 --device=/dev/dri \
 --group-add=video \
 --ipc=host \
 --cap-add=SYS_PTRACE \
 --security-opt seccomp=unconfined \
 --shm-size 8G \
 -v $PWD:/workspace \
 -e DISPLAY=$DISPLAY \
 genesis-amd

The examples will be accessible from /workspace/examples. Note: AMD users should use the ROCm (HIP) backend. This means you will need to call gs.init(backend=gs.amdgpu) to initialise Genesis.

Contributing to Genesis

The Genesis project is an open and collaborative effort. We welcome all forms of contributions from the community, including:

  • Pull requests for new features or bug fixes.
  • Bug reports through GitHub Issues.
  • Suggestions to improve Genesis's usability.

Refer to our contribution guide for more details.

Support

  • Report bugs or request features via GitHub Issues.
  • Join discussions or ask questions on GitHub Discussions.
  • Real-time chat: Discord.

License and Acknowledgments

The Genesis source code is licensed under Apache 2.0.

Genesis's development has been made possible thanks to these open-source projects:

Citation

If you use Genesis in your research, please consider citing:

bibtex
@article{
   genesis2026genesisworld,
   author = {Genesis AI Team},
   title = {The Role of Simulation in Scalable Robotics, Genesis World 1.0, and the Path Forward},
   journal = {Genesis AI Blog},
   month = {May},
   year = {2026},
   url = {https://www.genesis.ai/blog/the-role-of-simulation-in-scalable-robotics-genesis-world-10-and-the-path-forward},
}
bibtex
@misc{
  Genesis,
  author = {Genesis Authors},
  title = {Genesis: A Generative and Universal Physics Engine for Robotics and Beyond},
  month = {December},
  year = {2024},
  url = {https://github.com/Genesis-Embodied-AI/genesis-world}
}
<!-- Catalogue entries pruned from the Physics grid. Kept here as a reference so they can be reinstated later. The links and thumbnail paths are all still valid in the repo; just paste any pair of rows back into the Physics table. | [Rigid: grasp bottle](./examples/rigid/grasp_bottle.py) | [Rigid: collision pyramid](./examples/collision/pyramid.py) | [FEM: elastic dragon](./examples/elastic_dragon.py) | | | | | | [FEM: SAP fixed constraint](./examples/sap_coupling/fem_fixed_constraint.py) | [SPH: liquid](./examples/tutorials/sph_liquid.py) | [Coupler: grasp soft cube](./examples/coupling/grasp_soft_cube.py) | | | | | | [Coupler: cloth + rigid](./examples/coupling/cloth_attached_to_rigid.py) | [SAP: Franka grasp FEM sphere](./examples/sap_coupling/franka_grasp_fem_sphere.py) | [SAP: FEM sphere + cube](./examples/sap_coupling/fem_sphere_and_cube.py) | | | | | Pruned from Simulation Interface (same logic — labels/paths still valid): | [Entity name](./examples/tutorials/entity_name.py) | [Select rendered envs](./examples/tutorials/selecting_rendered_envs.py) | [GUI: keyboard teleop](./examples/keyboard_teleop.py) | | | | | | [Control franka](./examples/rigid/control_franka.py) | [Position control comparison](./examples/tutorials/position_control_comparison.py) | [IK + motion planning](./examples/tutorials/IK_motion_planning_grasp.py) | | | | | | [Close kinematic chain](./examples/rigid/closed_loop.py) | [Advanced: muscle](./examples/tutorials/advanced_muscle.py) | [Advanced: hybrid robot](./examples/tutorials/advanced_hybrid_robot.py) | | | | | -->