AGIBOT unveiled Genie Sim 3.0, a simulation platform designed to accelerate embodied AI development by integrating environment generation, data scalability, and evaluation. The Shanghai-based company enables users to create interactive 3D environments from text or images within minutes, bypassing traditional manual modeling. Genie Sim 3.0 supports multimodal data output aligned with real robot perception, enhancing training fidelity.
The platform includes a comprehensive benchmark evaluating five core robot capabilities: instruction following, spatial reasoning, manipulation skills, robustness, and sim-to-real transfer. It supports models like GO-2, Pi series, and GR00T series, providing systematic performance metrics in complex scenarios. Additionally, Genie Sim 3.0 integrates with the RLinf framework, enabling high-frequency physics simulation and massively parallel reinforcement learning to improve model convergence and control precision.
For AECM professionals, Genie Sim 3.0 offers a reusable infrastructure to develop and evaluate embodied AI applications efficiently, reducing data collection costs and improving scenario diversity. The platform's high-fidelity simulation and standardized evaluation can streamline robotics integration in manufacturing and construction projects, enhancing automation reliability and deployment speed.
AGIBOT plans to showcase Genie Sim 3.0 capabilities at the 2026 Robotics Summit & Expo in Boston on May 27-28. The platform's ongoing development aims to further bridge simulation and real-world robotic performance with expanded task suites and ecosystem compatibility.
Source: https://www.therobotreport.com/agibot-introduces-genie-sim-3-0-simulation-platform-embodied-ai/. Read the original story →
