The International Association for Automation and Robotics in Construction
Publications / ICRA 2024 Future of Construction Workshop Papers
Large Language Models for Robot Task Allocation
This study introduces an innovative framework that employs Large Language Models (LLMs) to enhance task allocation by seamlessly integrating construction robots and human users. The LLM contains key data about the task, such as agent capabilities, as well as details of the end goal to be achieved. An efficient allocation strategy is computed, balancing time efficiency and resource usage. By leveraging a Natural Language Processing interface, the system simplifies interactions with construction professionals and dynamically adjusts to unforeseen site conditions. Two LLM agents (a generator and a supervisor agent) are used concurrently to provide a more accurate task schedule. We test the proposed methodology with a simple scenario where the combination of two LLM agents provides a more accurate and logical schedule for the completion of a given task. The results highlight the significant potential of LLMs to transform operational tasks in construction, indicating a substantial step forward in aligning the industry with the latest developments in AI.