UC Berkeley and NVIDIA Announce Agentic Robotics Approach for Industrial Tasks

By ASR Newsdesk · 9 Nov 2026, 15:00 UTC

UC Berkeley and NVIDIA are introducing agentic robotics, a new AI-driven method for programming robots for industrial applications. This approach aims to overcome limitations of data-intensive imitation learning by focusing on human-readable code derived from a fundamental understanding of robot mechanics. The technique utilizes AI coding agents to decompose tasks, write, test, and modify software, leveraging existing robotics tools and LLMs. This method is presented as a way to address the data scaling challenges in robotics and improve reliability for industrial environments.