Publications / 2024 Proceedings of the 41st ISARC, Lille, France

Automating Construction Safety Inspections using Robots and Unsupervised Deep Domain Adaptation by Backpropagation

Vimal Velusamy Bharathi, Samuel Prieto, Borja Garcia de Soto, Jochen Teizer
Pages 855-862 (2024 Proceedings of the 41st ISARC, Lille, France, ISBN 978-0-6458322-1-1, ISSN 2413-5844)
Abstract:

Due to the dynamic aspect of construction sites, constant implementation and removal of safety equipment is a required practice. This leads to frequent manual and time-consuming inspections to make sure the safety measures are in place. There is the potential to automate the inspection process using robots and Deep Learning. Such an approach can save time and cost while improving safety. Using images collected by an Autonomous Ground Vehicle, a Deep Learning model with Domain Adaptation techniques is trained to detect and segment safety guardrails. The results of the model indicate a promising method to assist in automating site safety inspection that can make construction sites safer. Further work is necessary to validate this effort under more realistic and harsh construction site conditions.

Keywords: Construction equipment emissions, Digital Twin, Discrete Event Simulation, Site layout optimization, Non-road Mobile Machinery, Portable Emission Measurement System, Prediction and avoidance.