Genesis-Embodied-AI/genesis-world

Simulation platform for general-purpose robotics & embodied AI learning.

What it solves

Genesis World is a simulation platform designed for physical AI and robotics development. It provides a unified environment where researchers can simulate complex physical interactions—ranging from rigid bodies to fluids and deformable materials—while scaling from a single laptop to datacenter-grade GPUs.

How it works

The platform is structured into four primary layers:

  • Simulation Interface: A Pythonic API for managing assets (URDF, USD, etc.), sensors, controllers, and parallel environments.
  • Physics Engine: A multi-physics engine that integrates various solvers including Rigid, FEM, MPM, Particle (PBD/SPH), and IPC, all sharing a single scene and state.
  • Render Engine: Offers three rendering paths: Nyx (a dedicated robotics renderer), Luisa (a DSL ray tracer), and Pyrender (a rasterizer).
  • Compiler (Quadrants): A cross-platform compiler that lowers Python kernel code to multiple backends including CUDA, AMD ROCm, Apple Metal, Vulkan, x86, and ARM64.

Who it’s for

It is built for AI researchers and robotics engineers who need a scalable, high-performance simulation environment for training ML pipelines, generating synthetic data, or developing agentic simulations.

Highlights

  • Unified Multi-Physics: Supports rigid, soft, and fluid dynamics within one scene.
  • Cross-Platform Scaling: Runs on diverse hardware from laptops to GPU clusters via the Quadrants compiler.
  • Comprehensive Sensor Suite: Includes built-in support for depth cameras, IMUs, Lidar, tactile sensors, and temperature grids.
  • Photo-realistic Rendering: Integrated with the Nyx renderer for high-fidelity visual data.
  • Detailed Asset Support: Parses standard formats like URDF, MJCF, OBJ, GLB, and USD.

Related

  • Project
    galilai-group/stable-worldmodel

    A platform for reproducible world model research and evaluation that provides a unified interface for data collection, training, and model-predictive control evaluation.

  • Project
    SimWorld-AI/SimWorld

    An Unreal Engine 5-based simulation platform for developing and evaluating LLM/VLM agents in realistic physical and social environments.

  • Project
    chrxh/alien

    A GPU-accelerated artificial life simulation tool that uses a particle engine and neural networks to evolve digital organisms in a 2D physics environment.

  • Project
    AgibotTech/genie_sim

    A simulation platform for embodied intelligence that provides tools for 3D environment reconstruction, LLM-driven scene generation, and large-scale synthetic data collection for robot training.

  • Project
    NVIDIA/cosmos

    An open platform of omnimodal world models and tools for building Physical AI, enabling the joint processing and generation of text, vision, audio, and action sequences for robotics and autonomous systems.