What type of study is this?

September 5, 2025Open Access

Framework for LLM-Enabled Construction Robot Task Planning: Knowledge Base Preparation and Robot–LLM Dialogue for Interior Wall Painting

Key Points

LLMs can significantly enhance task planning in construction robots through adaptive dialogue, improving operational efficiency.
Using ChatGPT-4, the framework processes both structured BIM data and unstructured inputs for accurate robot instructions.
A prototype demonstrated that LLMs can execute tasks in real-time, reducing dependency on conventional hardcoded logic.
This approach indicates that LLMs have the potential to handle more complex operations beyond interior wall painting.

Abstract

Task planning for a construction robot requires systematically integrating diverse elements, such as building components, construction processes, user input, and robot software. Conventional robot programming complicates this by requiring precise entity naming, relationship definitions, unstructured language interpretation, and accurate action selection. Existing research has focused on isolated components, such as natural language processing, hardcoded data linkages, or BIM data extraction. We introduce a novel framework using an LLM as the cognitive core for autonomous construction robots, encompassing both data preparation and task planning phases. Leveraging OpenAI’s ChatGPT-4, we demonstrate how LLMs can process structured BIM data and unstructured human inputs to generate robot instructions. A prototype tested in a simulated environment with a mobile painting robot adaptively executed tasks through real-time dialogues with ChatGPT-4, reducing reliance on hardcoded logic. Results suggest that LLMs can serve as the cognitive core for construction robots, with potential for extension to more complex operations.

Read Full Paperexternally

AI से पूछें

Bookmark

View Full Paper