Task Order 5319
Traffic Operations Research


Evaluation of Pems to Improve the Congestion Monitoring Program

Pravin Varaiya and Jaimyoung Kwon
Institute of Transportation Studies
University of California, Berkeley

Bill McCullough
System Metrics Group, Inc.

Karl Petty
Berkeley Transportation Systems

Summary

Freeway congestion monitoring is critical in freeways operations and planning. Caltrans uses two approaches to monitor congestion. Most districts use tachometer vehicle runs. The alternative approach uses PeMS' HICOMP application.

The tachometer vehicle runs are resource intensive, requiring expensive personnel and vehicles. The PeMS-based approach is free, because it automatically processes loop data.

The crucial difference between the two approaches is that tachometer runs give a profile of individual trips, whereas fixed detectors give averages of traffic parameters at specific spatial locations. The implication for congestion monitoring of this difference is at present not known.

The goal of the proposed research is to determine

  1. How congestion and travel time estimates from PeMS for freeway sections equipped with loop detectors correlate with tachometer based estimates;
  2. The variation in accuracy of the PeMS-based estimates as a function of detector spacing;
  3. The variation in accuracy of the tachometer based estimates as a function of the time spacing of tachometer runs.
This research has the potential of saving valuable staff time and vehicle resources while delivering more accurate estimates of travel time and congestion needed for the Department's strategic objectives in improving mobility. Results of this study would also encourage full utilization of the existing detection infrastructure currently in place.

The research will develop, calibrate, and deliver separate models for the two alternatives: for tachometers, the model will take the time spacing of tachometer runs as the input variable; and for PeMS, it will take the detector spacing as the input variable. Both models also will take the traffic characteristic in the section as the common input variable. The empirical models will also produce measure of accuracy for the final estimates.

With these models, the two approaches will be compared in terms of the cost for achieving the same target accuracy. For PeMS, this cost is nearly zero. The cost of tachometers can be quantified as person-hours. The result can be used as a guideline to determine whether replacing tachometer runs with PeMS is recommended for a particular section. For un-instrumented freeways, a tachometer-based approach would still be used. Finally, a standardized congestion reporting protocol or methodology will be produced that can be implemented across Districts to accurately document congestion in a standardized fashion so the final state-wide summary will be coherent and complete over the entire network.