Tracking Divergence in Workers' Trajectory Patterns for Hazard Sensing in Construction

Abstract

In construction, hazard identification is a critical first step in safety management. To complement current hazard identification practices that rely on manual inspection, our previous study proposed an approach to identifying hazards by analyzing workers' collective behavioral responses-particularly the collective abnormality of workers' gait patterns as observed via foot-worn inertial measurement units. Alternatively, workers' moving trajectories represent behavioral responses that may contain important information about site conditions, since workers tend to change their walking path to avoid the risks posed by hazards. In this context, this paper presents an approach to quantitatively assess the level of divergence within workers' collective moving trajectories. The proposed approach first defines a representative trajectory from the quadratic polynomial regression-fitting of a given dataset and then assesses how divergent a moving trajectory is by quantifying the dissimilarity between each worker's moving trajectory and its representative trajectory. The results from a laboratory experiment simulating bricklaying activities demonstrates that such measurements of divergent trajectories can indicate the location of hazards.