Cost-effective Traffic and Roadway Data Collection Using Edge-based Comprehensive Sensing System: A Machine Learning Based Approach

The primary goal of this research is to develop a robust and cost-effective traffic-sensing and data-processing system including software algorithms and a hardware unit to address the current unequal and insufficient collection of traffic data. The software system must be compatible with the Washington State Department of Transportation (WSDOT) system and connected with the existing TMC equipment seamlessly. Moreover, the hardware system should be easy to install and maintain as well as a competitive cost. To achieve this goal, the research objective includes the following aspects: (1) Propose a machine learning (ML)-based approach for realizing cost-effective traffic and roadway data collection. ML technology has the capability to effectively learn and extract knowledge from large amounts of data. Specifically, deep-learning-based object detection, localization, classification, tracking, and counting will be utilized in this project due to its outstanding capability to learn deep representative features with high accuracy and efficiency. (2) Incorporate the algorithm in an edge-based comprehensive sensing system MUST. This means the whole data process will be finished on edge devices without the need to transfer data back to a server. One existing problem here is deploying the model efficiently on edge devices while considering computational restrictions. Also, the system should function well in locations with poor communication infrastructure or safety concerns due to geometry and volume constraints. (3) Collect and categorize important traffic information accurately and automatically. This information includes classified vehicle volumes (FHWA 13-bin classification), road surface conditions, visibility, and other relevant data. It is important to note that this system only focuses on the data collection of vehicles, not including pedestrians or cyclists. To sum up, the expected outcomes of this project are a traffic data sensing and processing system deployed in a mobile unit. This system should be capable to perform a 13-category vehicle classification and help researchers collect information where traditional methods cannot be used for safety or functional limitations. The collected traffic data will support data-driven decision making and more advanced technology development in related fields.

Language

• English

Project

• Status: Active
• Funding: $200000
• Contract Numbers:

  T1462-AI

• Sponsor Organizations:

  Washington State Department of Transportation

  Transportation Building
  Olympia, WA  United States  98504
• Managing Organizations:

  Washington State Department of Transportation

  Research Office
  P.O. Box 47372
  Olympia, WA  United States  98504
• Project Managers:

  Brodin, Doug

• Performing Organizations:

  University of Washington, Seattle

  Civil and Environmental Engineering Department
  201 More Hall, Box 352700
  Seattle, WA  United States  98195-2700
• Principal Investigators:

  Wang, Yinhai

• Start Date: 20230901
• Expected Completion Date: 20250831
• Actual Completion Date: 0
• USDOT Program: Highway and Transportation Data
• Subprogram: Highway Performance Management System (HPMS)V8

Subject/Index Terms

• TRT Terms: Data collection; Information processing; Machine learning; Traffic data
• Identifier Terms: Washington State Department of Transportation
• Subject Areas: Data and Information Technology; Highways; Operations and Traffic Management;

Filing Info

• Accession Number: 01902110
• Record Type: Research project
• Source Agency: Washington State Department of Transportation
• Contract Numbers: T1462-AI
• Files: RIP, STATEDOT
• Created Date: Dec 14 2023 12:47PM