Deep Reinforcement Learning-based Digital Twin for Risk Improved Decision Making in Transportation Construction

Each state in the U.S. including Louisiana is responsible to oversee an enormous number of construction and maintenance projects of transportation infrastructure systems such as highways, bridges, tunnels, and other infrastructure structures. In addition, with the increase in the need for constructing and repairing transportation systems, the state DOTs are daunted with the mammoth task of multiple projects and unexpected maintenance caused by recent natural disasters. This pandemic period has also been a strong driving force for DOTs to consider an advanced approach to remotely govern and support these multiple transportation projects. Organizing and overseeing a large number of transportation construction and maintenance projects that generally entail several miles of a worksite are a critical burden for each DOT. In addition, it requires manual monitoring of project or construction managers to identify a progress status, a work activity, and a safety issue in a job site. Because of the projected huge volume, complexity, significant impacts of future transportation infrastructure projects, it is evident that we are now facing a critical need to create a means of improving the results of work zone management and evaluating their impacts on our society. In addition, multiple work zones of a large-scale highway construction project usually have to be managed and monitored by a human effort on site, which is slow, inaccurate, and expensive. One primary problem in this situation is that it has been increasingly challenging for each DOT to consistently monitor progress of all projects in each state as well as efficiently evaluate work performance. With limited human resources and time, DOTs in Region 6 States have managed large-scale transportation construction and maintenance projects by a human inspection and recovered direct and indirect damages of transportation infrastructure systems caused from the recent natural disasters. Another critical issue is that this problem has prevented urban-level and integrated project management. Since it is not feasible to identify the status and the progress of numerous transportation infrastructure projects in real-time, DOTs cannot flexibly organize project resources and schedule according to diverse external factors including uncertainties in a worksite, mobility, natural disaster, and others. In particular, the lack of urban-level project and progress data is expected to be a critical obstacle for sharing real-time construction worksite information with autonomous vehicles and self-driving cars. The primary goal of this project is to identify the characteristics of the digital twin technology that are applicable to transportation construction and develop a conceptual framework of the prototype with a participatory sensing concept to improve the construction process monitoring, performance evaluation, and safety. The digital twin model incorporating the project information and schedules analyzes on-going activities and conducts thereby urban-level monitoring of all worksites. Recent research suggests that using only IoT sensors for capturing real-time data may be insufficient for entirely grasping the real-life situation. Involving participatory sensing along with IoT sensors for collecting real-time information can be a more efficient approach. Therefore, the aim of this study is to develop a conceptual framework of a digital prototype for managing and monitoring transportation construction projects using sound-based real-time data and participatory sensing along with the IoT sensors. This research involves sound-based data collection as it was found that audio-based approach can be used for activity identification of heavy-equipment with relatively high accuracy. Furthermore, sound-data as compared to image data is lightweight and can be easily processed, does not require a minimum level of illumination and thus is equally efficient during nighttime construction activities as daytime, and sound-sensors can capture data from unlimited angles unlike image-sensors.

• Supplemental Notes:
  • 22COLSU39

Language

• English

Project

• Status: Active
• Funding: $120000
• Contract Numbers:

  69A3551747106

• Sponsor Organizations:

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590
• Managing Organizations:

  Transportation Consortium of South-Central States (Tran-SET)

  Louisiana State University
  Baton Rouge, LA  United States  70803
• Project Managers:

  Mousa, Momen

• Performing Organizations:

  Louisiana State University, Baton Rouge

  P.O. Box 94245, Capitol Station
  Baton Rouge, LA  United States  70803
• Principal Investigators:

  Lee, Yong-Cheol

• Start Date: 20220401
• Expected Completion Date: 0
• Actual Completion Date: 0
• USDOT Program: University Transportation Centers Program

Subject/Index Terms

• TRT Terms: Asset management; Construction management; Construction projects; Data collection; Decision making; Infrastructure; Machine learning; Maintenance management; Monitoring; Prototypes; Real time information; Remote sensing; Technological innovations; Virtual reality; Work zones
• Subject Areas: Bridges and other structures; Construction; Data and Information Technology; Highways; Maintenance and Preservation; Planning and Forecasting;

Filing Info

• Accession Number: 01844764
• Record Type: Research project
• Source Agency: Transportation Consortium of South-Central States (Tran-SET)
• Contract Numbers: 69A3551747106
• Files: UTC, RIP
• Created Date: May 6 2022 11:07AM