Utilizing Archived Traffic Signal Performance Measures for Pedestrian Planning and Analysis

US Transportation Collection2020PDFTech ReportSingleton, Patrick A.Runa, FerdousyHumagain, PrasannaUtah State University. Department of Civil and Environmental EngineeringUtah. Department of Transportation. Research and Innovation DivisionUnited States. Department of Transportation. Federal Highway AdministrationUtah. Dept. of Transportation. Research DivisionUtahUnited StatesCluster analysisPedestrian movementPedestrian trafficRegression analysisSignalized intersectionsTime series analysisTraffic dataTraffic volumeVideoTraffic actuated controllersPedestriansPedestrian countsPedestrian detectorsTraffic signalsActuated traffic signal controllersWalkingPedestrian actuated controllersUT-20.1719-8237Final Report, Sep 2018 to Jul 2020The overall goal of this project was to explore the use of continuous pedestrian traffic signal data from the Automated Traffic Signal Performance Measures (ATSPM) system to estimate pedestrian volumes at signalized intersections. Our objectives were to identify patterns of pedestrian activity at signalized intersections, develop methods to estimate pedestrian crossing volumes from signal data, and create a prototype visualization. Using one year of data from 1,522 Utah traffic signals, we applied time series clustering across two new \u201cpedestrian activity metrics\u201d and identified seven distinct patterns of hourly and weekly pedestrian activity. These seven typologies varied by the magnitude (high, medium, and low) and the number (one or two) of weekday peak hours. Based on these typologies, we randomly selected 90 Utah signals, used UDOT traffic cameras to record over 10,000 hours of video, and manually counted almost 175,000 pedestrians crossing at the intersections. Using processed hourly pedestrian actuations and detections from ATSPM data, we estimated five non-linear regression models (segmented by pedestrian activity, cycle length, and pedestrian recall) using pedestrian signal data to predict hourly pedestrian crossing volumes. Overall, our estimates were strongly correlated with observed volumes (0.84) and had a low error (+/- 3.0 on average). These results\u2014which use orders of magnitude more data than had previously been assembled on this topic\u2014demonstrate the validity of using pedestrian data from traffic signals to estimate levels of pedestrian activity. We also developed a prototype dashboard to interactively visualize pedestrian signal data.109