State Space Exploration with Large Language Models for Human-Robot Cooperation in Mechanical Assembly

In the realm of human-robot cooperation (HRC) for mechanical assembly, determining the task allocation between human operators and robots is crucial. Traditionally, this requires extensive pre-planning of the product's state space. We propose a novel system utilizing a Large Language Model (LLM) to dynamically reason and determine the next assembly state based on real-time product description. A camera monitors the assembly process, and the captured data is pre-processed and fed to a LLM, which predicts the next assembly state and subsequently instructs the robot on its task. This approach enables a more flexible and adaptive assembly process without the need for exhaustive pre-planning.