Task Planning for a Factory Robot Using Large Language Model

Abstract

In recent years, automation has significantly advanced the automobile manufacturing industry. However, many tasks still involve human intervention, so there is a demand for the development of robots to support workers. Additionally, as the human-centric approach, such as Industry 5.0, is gaining attention, it is expected that support robots like these will become necessary in the future. This study aims to develop a system that can support workers by utilizing robots that anyone can easily use and flexibly respond to various tasks. This system adopts a large language model (LLM) for work planning and generates tasks that robots can execute by making bidirectional and interactive suggestions and modifications through natural language dialogue in response to human demands, aiming to improve further productivity and the working environment in automobile manufacturing factories. The proposed system was tested in a simulated factory environment and then the performance was confirmed in an actual factory setting. And it was confirmed that various tasks can be executed by robot through work planning and dialogue with the LLM.