Multi-Stage Algorithm for Detection-Error Identification and Data Screening

In state of Utah, various types of traffic sensors, such as loop and radar detectors, have been widely deployed on freeways to support advanced traffic management systems. However, through previous efforts of using historical data, the research team has observed many detection errors of flows and speeds. This might be caused by the lack of detector maintenance and re-calibrations. Consequently, it can prevent effective applications of many control systems that depend on traffic data as input. Therefore, it is essential to develop a well-designed screening algorithm that is capable of identifying the potential detection errors and highlighting the stations that should have maintenance priorities. Utah Department of Transportation (UDOT) Traffic Operation Center (TOC) is managing several databases, such as Freeway Performance Metrics, Performance Measurement System (PeMS) and iPeMs, for traffic information storage. Engineers can get access to those databases to retrieve historical traffic data. However, no screening algorithm has yet been implemented to evaluate the data quality and identify the detection errors in the database. In this project, the research team aims to develop a multi-stage data screening tool to fulfill the following goals: (1) Evaluate data qualities of those databases and identify the potential detection errors; (2) Perform further statistical analysis to confirm the detection errors; (3) Conduct in-depth review of stations with potential detection errors and assess their needs for maintenance or re-calibration; and (4) Identify high speed locations that require speed reinforcement using validated dataset. The multi-stage screening algorithm will include the following key step: (1) data completeness check and missing data identification; (f2) multiple database cross-verification; (3) single-variable threshold check, (4) multi-variable threshold check; (5) statistical test with temporal and spatial information comparisons; and (6) long-term data analysis to identify malfunctioning detectors. The algorithm will categorize the data into three groups: no error, error, and potential error with probability. Average Effective Vehicle Length (AEVL), which can be computed with occupancy rate, flow rate, and speed, will serve as a key measurement for data screening. Notably, either temporally or permanently malfunctioning sensors can produce detection errors in practice. Hence, it might be difficult to distinguish them based on a short time period of data. In this project, the research team will develop a methodology that can conduct in-depth data reviews of those stations with potential detection-errors. Such effort will assist UDOT to quickly locate and prioritize the ones require immediate care.

  • Supplemental Notes:
    • Contract to a Performing Organization has not yet been awarded.

Project

  • Status: Active
  • Sponsor Organizations:

    Utah Department of Transportation

    4501 South 2700 West
    Project Development
    Salt Lake City, UT  United States  84114-8380
  • Start Date: 20190124
  • Expected Completion Date: 0
  • Actual Completion Date: 0

Subject/Index Terms

Filing Info

  • Accession Number: 01703044
  • Record Type: Research project
  • Source Agency: Utah Department of Transportation
  • Files: RiP, STATEDOT
  • Created Date: Apr 25 2019 8:22PM