AgibotTech/genie_sim

What it solves

Genie Sim is a comprehensive simulation platform designed to accelerate the development of embodied intelligence (robotics). It reduces the reliance on physical hardware by providing tools for high-fidelity environment reconstruction, synthetic data collection, and automated evaluation of robot policies.

How it works

The platform combines 3D reconstruction (using 3D Gaussian Splatting) and visual generative models to create photorealistic simulation environments. It uses LLMs to generate diverse scenes from natural language prompts and provides a real-time (RT) engine integrated with ROS 2 for interactive simulation. The system includes a benchmark suite with over 100,000 scenarios to evaluate model capabilities in manipulation and cognition.

Who it’s for

It is built for robotics developers and AI researchers focusing on embodied AI, specifically those needing to generate large-scale synthetic datasets or validate robot policies before deploying them to real-world hardware.

Highlights

• LLM-Driven Scene Generation: Create complex simulation scenes instantly via conversational interaction.
• Agent-Ready SKILLs: Machine-readable recipes that allow AI coding agents (like Claude Code) to operate the simulator without human intervention.
• High-Fidelity Assets: A library of over 5,000 validated 3D assets across retail, industrial, catering, home, and office environments.
• Zero-Shot Sim-to-Real: Synthetic data generation that enables models to transfer capabilities to real robots with high success rates.
• VLM-based Auto-Evaluation: An automated system using Vision-Language Models to profile a model's manipulation and cognitive skills.