AI agents can now write and test robot code from scratch, skipping months of manual work.

Researchers from UC Berkeley and NVIDIA have released an open-source toolset called RoboSimStudio that enables teams of AI coding agents to generate, simulate, and refine programs for industrial robots. The system uses large language models to write code but introduces a novel agentic approach where multiple AI agents collaboratively manage the coding, testing, and deployment process, moving beyond simple 'vibe coding' for automation.

This represents a fundamental shift in industrial robotics programming, which traditionally requires weeks to months of manual work after hardware installation, often under tight project deadlines. The toolset could drastically reduce the time and specialized expertise needed to program complex robotic workcells, making robotics more accessible and accelerating deployment cycles. The key breakthrough is enabling AI to create functional robot programs without relying on massive, pre-collected training datasets.