Improved Error Detection Using Prediction Techniques and Video Imaging

This research project evaluated an algorithm developed in the previous project (Nihan et al., Detector Data Validity), and developed a new data error detection algorithm by employing a video imaging data collection technology called Autoscope™. This new algorithm was calibrated with data from the Seattle metropolitan area. It helps to determine the reliability of 20-second loop detector data that are used for the operation of the ramp metering system. Both the existing and the new algorithms were tested for their effectiveness with an extensive data set that contains manually simulated erroneous data. The test data were collected from various locations on I-5 that covered different characteristics such as lane type, lane configuration, and geometrics. While both algorithms were effective in screening out hanging-off errors, chattering, and spurious pulses, the new algorithm provides a much more effective detection for hanging-on errors, especially in congested conditions.

The principal findings and recommendations of this research were as follows:

  1. The Autoscope™ data collection results were checked against itself for internal consistency and tested against manual counts for accuracy. We have found that the results were consistent with the developer's claim of an accuracy level of 92.18 percent to 98.32 percent for traffic counts.
  2. The new error detection algorithm resulting from this project showed a marked improvement over the original one, especially in screening out the hanging-on errors that occur in congested conditions.
  3. The feasible region of volume/occupancy data fell within two parabolic envelopes, substantiating the traditional understanding of this relationship.
  4. Except for chattering data and spurious pulses (which are treated as erroneous data), the other error flags can denote either detector malfunctions, or the existence of some "abnormal" traffic pattern, such as that caused by an incident. Therefore, the recognition of the erroneous data's location can possibly help to identify incidents happening in congested traffic. However, since this particular function has not been tested in this project, it may be an issue worthy of future investigation.
  5. A preliminary investigation of the relationship between vehicle length and the g-value was done; and empirical 20-second data supported the theoretical understanding of this relationship. It was recommended that further studies with different time slices be done to further investigate this relationship.
  6. The new error detection algorithm can be implemented in the WSDOT control system in the Seattle I-5 corridor. It will improve the integrity of the loop data, and hence, improve ramp control and freeway operation.

The Autoscope™ system can be used for algorithm development and for calibration of other facilities, such as HOV lanes. It can also be used for real-time data collection, analysis, and traffic control particularly at construction sites, where detection loop operations are usually interrupted.

Publication Date: 
Thursday, June 1, 1995
Publication Number: 
WA-RD 386.1
Last modified: 
10/12/2016 - 15:42
Nancy L. Nihan, Morgan Wong.
Washington State Transportation Center (TRAC)
Number of Pages: 
Accuracy, Algorithms, Fault location, Freeway operations, Improvements, Incident detection, Loop detectors, Ramp metering, Video imaging detectors, Work zone traffic control.