Experience assessing the accuracy and performance red-light extension (RLE) systems in Oregon.
Portland, Oregon, United States
Smart Red Clearance Extensions to Reduce Red-Light Running Crashes
Summary Information
This research evaluated the implementation of red-light extension (RLE) systems to reduce red-light running (RLR) and improve safety in Portland. An Initial field study was conducted to assess driver behavior and RLR characteristics at five intersections. The data collected were used to develop and calibrate a simulation model that could be used to evaluate alternative RLE solutions and assess detection and timing strategies at the US30 and Cornelius Pass Road intersection where a large number of RLR events and high operating speeds were common. Over the course of 24 hours of transcribed data at this intersection, 24 RLR vehicles were observed. On average, RLR vehicles proceeded through the intersection when they were approximately 64 ft from the stop line (0.821 sec) and when they were traveling at speeds considerably above the posted speed limit (58.4 mi/hr).
METHODOLOGY
A model of the US30–Cornelius Pass intersection was developed the using VISSIM-6 model. The intersection operates with a 2070 controller and NWS Voyage™ firmware and uses the NWS Voyage™ RLE function on through movements along US30 and Cornelius Pass Road for the left turn movement. RLE events are triggered by loop detectors located downstream from the stop line.
A number of different operating scenarios were modeled to compare the downstream detection (DD) strategy currently deployed with an alternative smart upstream speed-conditional detection (SUSCD) that detects vehicle trajectories upstream from the stop line. Performance data were compiled for each alternative.
- Existing DD RLE system that uses a single loop detector (per lane) located downstream of the intersection stop line.
- SUSCD modeled with detection 215 ft upstream
- SUSCD modeled with detection 475 ft upstream.
FINDINGS
Although the DD alternative provided higher accuracy than the SUSCD system at either 215 or 475 ft, the predictive SUSCD system consistently provided a greater degree of safety for vehicles that received a RLE with no significant increase in intersection delay.
Detection rates were high for the DD alternative because no prediction was made; the vehicle was already in the intersection when it was detected. Although the SUSCD system at 215 or 475 ft had a higher likelihood of false prediction of a RLR vehicle compared to the DD system, the SUSCD systems also introduced the potential for providing more robust RLE. An examination of time space diagrams (TSDs) showed improved relationships between the vehicle trajectories, RLE events, and conflicting movements when the SUSCD systems were used.
