From Retrieval to Action: GraphRAG Meets Physical AI

What if a robot could take a plain-English command and reason its way to a valid, ordered plan of action? In this session, we explore how GraphRAG can serve as the thinking and planning backbone for Physical AI, the systems that perceive, reason, and act in the real world. We'll break down how a robot's knowledge, its skills, the objects around it, and the rules that govern them can be modeled as a knowledge graph, and how multi-hop retrieval turns that graph into correctly-ordered, executable plans that respect real-world constraints. Along the way, we'll cover the architecture that connects a graph-powered planning layer to a live robot, using Arango's Contextual Data Platform and the Agentic AI Suite with GraphRAG. Whether you work with RAG, robotics, or neither, you'll leave with a clear picture of how retrieval becomes action.