Exploring the Quantitative Impact of Localization Accuracy on Localization-Based Safety Monitoring’s Performance on a Construction Jobsite

Abstract

The construction industry continues to be one of the industries with the highest accident rate. With the recent advancement of computing and sensing technologies, more efforts have been made in the field of context-aware autonomous job site monitoring, which can automatically track and evaluate the worker’s behavior and job site conditions in real time to issue warnings for unsafe behaviors and conditions on sites. The location of an entity (e.g., worker, equipment, or material) is one of the core information for context-aware management. Therefore, a large number of recent studies focus on exploring different approaches to improve the localization technologies’ accuracy, robustness, and convenience of deployment. However, systematic knowledge is missing on how the localization algorithm’s accuracy affects the construction safety monitoring’s performance. To quantify the impact in this paper, first, a localization system accuracy model and three safety clearance models are proposed as the base. Second, a Monte Carlo simulation model is designed as an abstract representation for localization systems. By conducting the simulation, the effects of the localization errors on safety monitoring performance are quantitatively analyzed. A case of a tower crane scenario is investigated as a demonstration. With the simulation data, a regression model is established, indicating that the precision and recall of the location-based safety decision making are not only related to the accuracy of the localization system but also related to the size of the safety zones. With the proposed model, the impacts of a given localization system on the location-based safety decision making can be quantified without the actual deployment, offering a more straightforward evaluation tool of the localization system.