Reproduce “Spatial and Semantic Reasoning for LLM-Driven Robot Navigation via MCP” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jungsoo Lee, Jaegyun Park, Wansoo Kim
The authors present reusable MCP tools that give large language models navigation context without modifying a robot’s ROS navigation stack. The tools convert occupancy grids into metric, pose-aware images and store natural-language observations with exploration waypoint poses. In a simulated warehouse, GPT-5.5 achieved 99.78% map coverage, 28/30 room-navigation successes, and 30/30 fire-extinguisher waypoint successes, under annotated-waypoint limitations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.