Integrating LLMs and Digital Twins for Adaptive Multi-Robot Task Allocation in Construction

Multi-robot systems are emerging as a promising solution to the growing demand for productivity and safety across industrial sectors. However, effectively coordinating multiple robots in dynamic and uncertain environments, such as construction sites, remains a challenge, particularly due to unpredictable factors like material delays, unexpected site conditions, and weather-induced disruptions. To address these challenges, this study proposes an adaptive task allocation framework that strategically leverages the synergistic potential of Digital Twins, Integer Programming (IP), and Large Language Models (LLMs). The multi-robot task allocation problem is formally modeled using IP, considering task dependencies, robot heterogeneity, scheduling constraints, and replanning needs. A narrative-driven schedule adaptation mechanism is introduced, where unstructured natural language inputs are interpreted by an LLM to autonomously update optimization constraints, enabling human-in-the-loop flexibility without manual programming. A digital twin system enables real-time synchronization between physical operations and their digital representations and serves as the user interface, creating a closed-loop feedback mechanism that keeps the system responsive to site changes. A case study demonstrates both the computational efficiency of the optimization algorithm and the reasoning performance of LLMs, achieving over 97% accuracy in constraint and parameter extraction. The results confirm the practicality, adaptability, and cross-domain applicability of the proposed methods. Note to Practitioners—Construction projects often face sudden disruptions, such as changing priorities, late material arrivals, or weather delays, that can affect planned robot operations. This work presents a practical approach through a real-time site digital twin, connected to a language-based interface that understands plain human instructions. The system is integrated with a scheduling module that can allocate and replan tasks based on human instructions. This enables site managers to adapt robot assignments instantly without modifying the code manually, preserving operational efficiency and minimizing downtime. While the current simulation setup simplifies certain aspects of real-world construction scenarios, the framework is designed for straightforward integration with existing multi-robot and human-robot collaborative construction systems, making it well-positioned for near-term industry adoption. Beyond construction, it can benefit manufacturing, logistics, and other domains requiring flexible multi-agent coordination.