RESEARCH · RESEARCH · #1086
Stanford/Caltech team connects GPT-6 Astra directly to a Unitree G1 to tidy unfamiliar kitchens
Researchers at Stanford and Caltech built HomeBody, a system that lets a Unitree G1 robot explore an unfamiliar kitchen, build a digital twin in Nvidia Isaac Sim, and use GPT-6 Astra as a vision-language model that calls an extensible skill library to grasp, navigate, and open drawers so it can tidy and fetch items. The setup drops the usual trained control layer—Astra plans and self-corrects steps using spatial memory, but the team reported practical limits including latency, overheating finger servos, and high compute costs; the code is available on GitHub.
KEY POINTS
- Researchers at Stanford and Caltech built HomeBody, a system that lets a Unitree G1 robot explore an unfamiliar kitchen, build a digital twin in Nvidia Isaac Sim, and use GPT-6 Astra as a vision-language model that calls an extensible skill library to grasp, navigate, and open drawers so it can tidy and fetch items.
- The setup drops the usual trained control layer—Astra plans and self-corrects steps using spatial memory, but the team reported practical limits including latency, overheating finger servos, and high compute costs; the code is available on GitHub.
- This shows a high-capability LLM (GPT-6 Astra) can be used to directly control robotic skills and planning without a separate trained policy, demonstrating new possibilities and clear safety, hardware, and compute trade-offs.
WHY IT MATTERS
This shows a high-capability LLM (GPT-6 Astra) can be used to directly control robotic skills and planning without a separate trained policy, demonstrating new possibilities and clear safety, hardware, and compute trade-offs.