This paper presents an LLM based safety aware navigation system for mobile robots operating near humans. The robot detects people and obstacles using camera–LiDAR sensing and computes a Human Proximity Safety Score (HPSS) to identify risk events. Wh...
This paper presents an LLM based safety aware navigation system for mobile robots operating near humans. The robot detects people and obstacles using camera–LiDAR sensing and computes a Human Proximity Safety Score (HPSS) to identify risk events. When a risk event occurs, structured sensor information is passed to a lightweight LLM, which interprets why the situation is dangerous and outputs adjusted navigation parameters such as reduced speed and increased safety distance. Experiments in a warehouse simulation with varying crowd levels show that the proposed system reduces unnecessary stops and maintains safer distances compared to Nav2, demonstrating improved safety and stability in human-centered environments.