AUTOMATED SCHEDULING METHOD FOR REDUCING SPATIAL-TEMPORAL CONFLICT SAFETY RISKS, USING ML AND BIM

Abstract

SUMMARY: Construction sites face inherent risks from overlapping activities in confined spaces and require advanced solutions to manage spatial-temporal conflicts. Surpassing previously developed static BIM tools, this study introduces an automated approach that revolutionizes job site safety planning by dynamically assessing conflict risks between evolving workspaces. The method integrates empirical workspace geometry with machine learning—using Support Vector Machine regression to predict worker presence—and BIM-driven dynamic conflict analysis, which reflects real-time changes in workspace sizes and team movements. By generating safety score matrices through pairwise risk assessments, it quantifies conflict types (physical impacts, proximity risks, workflow disruptions) and enables real-time scenario comparisons via Python-based evaluation. Adjustable parameters allow customization for team sizes, workspace allocations, and pandemic-specific adaptations such as social distancing. Case studies show that the system effectively identifies high-risk periods, compares different work sequences, and makes schedules without sacrificing productivity. Unlike earlier clash detection methods that only compared static models, this framework provides actionable safety metrics to proactively respond to conflict risks. Designed for scalability, the presented method manages computational demands in complex projects. This advancement represents a paradigm shift in construction safety, blending predictive analytics with practical adaptability to protect workers and streamline operations.