AI and IoT-driven sensor technologies for real-time monitoring and control in construction

다국어 초록 (Multilingual Abstract)

The construction industry has not benefited greatly from current research on AI and IoT-driven sensor technologies, which has mostly concentrated on smart cities, manufacturing, and healthcare. Research currently being conducted tends to focus more on data gathering and simple automation than on incorporating sophisticated AI for real-time decision-making and predictive analytics. There is a research gap because stronger systems are required to manage the unstable and dynamic environment found on building sites. By creating a cutting-edge AI and IoT-based system specifically designed for real-time monitoring and control in the construction industry, our study fills this gap. This framework offers a considerable advantage over present technologies by improving not just data accuracy and sensor dependability but also safety, optimization of resource allocation, and predictive maintenance. The purpose of the project was to develop sensor technologies powered by AI and IoT for real-time construction monitoring and control. There were 1,198 street images and 330,165 individuals that comprise the ShanghaiTech dataset were gathered from the campus of Shanghai Jiao Tong University. In order to ensure consistency and reliability, the raw data prior to processing has been adjusted using the min-max normalization technique. We presented the Deep Deterministic Policy Gradient Algorithm with support vector machine (DDPGA-SVM) to provide real-time monitoring and control of construction-related sensors powered by AI and the IoT-driven. To evaluate the suggested solution works in terms of accuracy, prediction rate, loss function, F1-score and Cohen kappa score. As a result, real-time monitoring and control in construction demonstrated by the suggested superior performance over other similar models in terms of accuracy (99%), MAE (28%), F1-score (90), and recall (96) and loss function achieving 80% in training and 92% in validation.