Automated optimization of visual sensor deployment for construction sites

초록

Continuous visual data acquisition is essential for synchronizing physical construction sites with digital twins to ensure safety, quality, and productivity. However, conventional deployments often rely on subjective judgment and fail to account for complex physical constraints like topography, leading to blind spots and inefficient deployment. To address this, the paper proposes an automated, computer-aided decision support framework for objective and accurate visual sensor deployment. The framework integrates a grid-based spatial model with metaheuristics (Simulated Annealing and Late Acceptance Hill Climbing) and was validated across three real-world construction sites. Results demonstrated that the Precision-oriented Model (SA-based) achieved 99.15% coverage using only seven sensors, while the Runtime-oriented Model (LAHC-based) generated viable redeployment layouts within one minute. The paper translates irregular physical constraints into a quantitative optimization problem and provides an automated deployment framework for achieving geometric visibility across designated monitoring areas, thereby supporting continuous visual data acquisition for construction site digital twins.

키워드