Description
Abstract— Heritage and vernacular construction methods (e.g., interlocking joints and low-mortar masonry) are
increasingly revisited for sustainable building; however, their operational impacts on modern project performance are
rarely quantified under reproducible AI/ML protocols. This paper proposes an Ancient Inspired Multimodal Construction
Optimization (AIMCO) framework that encodes “heritage-inspired” design choices (interlocking masonry, mortar
reduction, and geometry proxies) alongside modern site signals (quantities, labour rates, critical path schedule logs, and
optional UAV/drone imagery) to predict cost and time overruns and to recommend Pareto efficient planning options.
Using a dataset schema aligned to 5D BIM cost/schedule attributes, AIMCO trains and compares linear regression, random
forests, gradient boosting (XGBoost), and neural networks for cost and duration prediction. When imagery is available, a
computer vision module estimates physical progress and detects wall defects, feeding back to the predictive models and
schedule deviation alarms. In the absence of field data, we provide a clearly labeled synthetic pilot dataset and illustrative
experiments demonstrating how heritage features and UAV-derived progress indices can improve error metrics relative to
CPM-only baselines. The manuscript includes variables, constraints, algorithm pseudocode, evaluation metrics
(MAE/RMSE/MAPE, F1, mIoU), and a reproducibility checklist to support replication on real projects.
Index Terms— Construction management, cost estimation, schedule optimization, building information modeling (BIM),
unmanned aerial vehicles (UAV), computer vision, interlocking masonry, life cycle assessment (LCA), machine learning,
multi-objective optimization.
Reviews
There are no reviews yet.