Utah, United States
Evaluating Ramp Meter Wait Time in Utah
Summary Information
Ramp metering allows freeways to operate with improved traffic flow by controlling the entry of ramp traffic onto freeways, particularly around peak hours. Researchers in Utah developed algorithms to predict wait times at metered on-ramps using data from existing loop detectors on the ramps. Five days of data were collected during the afternoon peak period at four on-ramps to I-15 throughout Davis, Salt Lake, and Utah counties. The algorithms used the data from the loop-detectors including occupancy, flow rate, density, and metering rate in 60-second increments to predict queue length. Queue lengths could then be converted based on the metering rate to provide travelers with estimated wait times. Models explored included a traffic conservation model and several variations of a Kalman filter model.
Lessons Learned
- Consider detector data accuracy rate in selecting algorithms. Researchers determined that loop detectors were not entirely accurate in their reporting of volume and occupancy data. The volumes entering and exiting the ramp were often several vehicles off from what the field-collected data showed. The imperfect data provided by the detectors led to inaccurate queue length and wait time estimates if using the conservation model.
- Use data filters to improve the accuracy of queue length and wait time predictions. A filtering model such as Kalman Filtering should be used to utilize other detector and ramp data to compensate for inaccuracy in individual detector counts.
- Consider the ease of implementation when adopting a wait time estimate model across all metered on-ramps. Both the heuristic and fixed Kalman constant model generally yielded wait time estimates within 45 seconds of the observed wait time and queue length estimates within 10 vehicles of the observed queue length. Although the heuristic model performed slightly better overall, it was more complex to implement. The fixed Kalman constant model was simple to implement and can be automated to provide wait time estimates at any metered on-ramp throughout the state across varied traffic conditions.
