This paper describes an evaluation of the safety impacts of a freeway ramp meter installed in Glasgow, Scotland. Glasgow is a city of approximately 625,000, within an urban area of 1.7 million people. The city is currently undertaking numerous projects to improve the safety of the roadway system. During the five-year period prior to the date of this study (1998), the city experienced a 52 percent reduction in the number of fatal accidents and a 19 percent reduction in injury accidents. Several other ITS projects are being undertaken in the area, including variable message sign assisted park and ride, traffic control strategies that assist public transportation at signalized intersections, and an integrated control strategy incorporating freeways and urban streets. The ramp metering project evaluated in the paper is a part of the integrated control project.

The ramp meter evaluated during this project, on an entrance ramp to the M8 motorway, was implemented to address a PM peak congestion problem. The meter is activated by heavy traffic volume on the freeway, however, and could be activated at anytime. During periods of ramp meter operation or, in extreme cases, ramp closure, traffic signal timings on the adjacent arterial adjust to accommodate additional volume traveling to the next downstream freeway entrance.

The project used video from two traffic cameras, one upstream of the ramp and one downstream, to assess the safety improvement gained through the ramp meter. An automated analysis system identified corner features in the video image and tracked their motion through the frame. Movement of these corner features allowed the determination of percentage of vehicles merging early, the percentage of vehicles changing lanes in the two lanes adjacent to the ramp within the area immediately prior to the merge area, and the mean deceleration of vehicles entering the merge area. Due to the positioning of the cameras, aggregating the corner features to identify individual vehicles was not possible.

Findings from the study indicate that the number of vehicles merging too early decreased after the implementation of ramp metering, from 35 percent to 25 percent. The study also found that the amount of lane changing in the two lanes adjacent to the ramp appears to have dropped. Prior to implementation, the percentage of vehicles changing lanes in this area ranged from 6 to 15 percent, while after ramp metering began this number ranged from 1 to 8 prcent. The swaying of the camera towers and their distance from the merge area lead to artificially high measured speeds for slow moving vehicles. This prevented quantitative measurement of vehicle deceleration. Qualitative analysis indicates that mean speed changes during ramp metering appear smaller than prior to implementation, but the authors stress that additional analysis is required for a quantitative determination.