Integrated 4D/5D Digital-Twin Framework for Cost Estimation and Probabilistic Schedule Control: A Texas Mid-Rise Case Study (2511.15711v1)

Abstract: Persistent cost and schedule overruns in U.S. building projects expose limitations of conventional, document-based estimating and deterministic Critical Path Method (CPM) scheduling, which remain inflexible under uncertainty and lag dynamic field conditions. This study presents an integrated 4D/5D digital-twin framework unifying Building Information Modeling (BIM), NLP, reality capture, computer vision, Bayesian risk modeling, and deep reinforcement learning (DRL) for construction cost and schedule control. The system automates project-control functions by: (a) mapping contract documents to standardized cost items using transformer-based NLP (0.883 weighted F1 score); (b) aligning photogrammetry and LiDAR data with BIM to compute earned value; (c) deriving real-time activity completion from site imagery (0.891 micro accuracy); (d) updating probabilistic CPM forecasts via Bayesian inference and Monte Carlo simulation; (e) using DRL for adaptive resource allocation (75% adoption rate); and (f) providing 4D/5D decision sandbox for predictive analysis. A Texas mid-rise case study demonstrates localized cost adjustment using RSMeans City Cost Index and Bureau of Labor Statistics wage data. Results show 43% reduction in estimating labor, 6% overtime reduction (91 hours), and project completion matching P50 probabilistic forecast of 128 days, confirming improved estimation accuracy and responsiveness.