Current Research

LLM-based construction scheduling and control agent.
Construction schedules operate under high variability—across trades, crews, work zones, materials, equipment, and subcontractors—and plans often change in response to on-site conditions. Yet schedule information is still frequently managed through documents and spreadsheets, making it difficult to keep plans consistently updated when changes occur, especially under constraints in specialized planning resources and time. Even when schedules are shared, differences in role-specific interpretation and execution constraints across subcontractors can lead to misalignment between the plan and field execution.

To address this, my research focuses on structuring and formalizing field scheduling information into machine-interpretable representations, enabling consistent schedule updates when disruptions such as delays or scope changes arise. Beyond this foundation, I aim to analyze delay propagation and generate alternative response options, comparing them through simulation-based evaluation to support practical decision-making. Ultimately, the goal is to extend the agent’s capabilities to the execution layer—supporting task coordination, responsibility assignment, and work allocation in the field.