Research Papers

Midtown Atlanta Case Study - New Data Collection Techniques in Automated Pedestrian Safety Analysis

Version 1
Date added June 18, 2019
Downloaded 0 times/fois
Category 2019 CARSP XXIX Calgary
Tags Research and Evaluation, Session 5A
Author/Auteur St-Aubin
Stream/Volet Research and Evaluation
Slidedeck Presentation Only:



Background/Context: Atlanta Midtown Alliance needed an effective pedestrian safety evaluation technique that quickly identifies performance changes in high-risk behaviour to effectively quantify pedestrian safety before and after the integration of an all-WALK pedestrian phase into the signal cycle at the busy intersection of Spring St. and 5th St. The conflict analysis was required to meet the increased demand to validate safety performance measurements with emphasis on Vision Zero.

Aims/Objectives: The intersection of Spring Street and 5th Street is a busy, high-risk intersection with pedestrian signals in midtown Atlanta near the Georgia Institute of Technology. There was pressure within the community to make interventions to reduce pedestrian injuries and collisions. A need for a safer crossing system was identified in "complete street" project to meet the sustainable mobility needs of all road users.

In January 2018, an all-WALK pedestrian phase was integrated into the signal cycle at the intersection of Spring Street and 5th Street in midtown Atlanta. The project deployment team, which included Midtown Alliance, Aecom and Jacobs, decided to use BriskLUMINA as their on-demand automated video-based road conflict analysis solution. Methods/Targets: BriskLUMINA was implemented with temporary cameras to track and classify all individual road users from video data, and automatically detect near-misses and other dangerous interactions between them that are indicative of future potential collisions.

Video data was collected at this intersection before and after the implementation of the all-WALK phase. The AI-driven software platform evaluated the potential conflicts between users of the all-WALK phase on pedestrian safety at the intersection. Using traffic cameras, all road users can be observed simultaneously.

For conflict analysis, the main safety measures to detect and identify collision risk are Time to Collision (TTC), Post Encroachment Time (PET) and vehicle speed, between any groups of scenario movements among vehicles, pedestrians, or cyclists. A potential conflict is identified when the PET is less than 3 seconds. Six conflicting movement scenarios were analyzed in 60 hours of both the before and after videos. Video data collection for both before and after were; morning period 7-9am, midday period 11am-1pm and afternoon period 4-6pm

Other considerations when implementing a new data-driven collection technique for crosswalk interventions:

- Before and after video resolution quality and camera angle proximity
- Defining selected intersection conflict scenarios and locations
- Pedestrian and driver signalized cycle compliance
- Education and enforcement
- Each road user is classified by mode of transportation or vehicle classification, and is grouped into a movement.

Results/Activities: The results indicate that the all-WALK pedestrian phase provides a significant benefit to pedestrians. Overall, the percentage of pedestrians involved in a potential conflict (PET < 3 seconds) at the intersection of Spring Street and 5th Street reduced from 4% before to 1% after intervention, representing a 75% improvement.

In addition, the percentage of potential conflicts where the pedestrian arrives to the potential conflict point first, decreased from 57% to 30% (approximately 47% reduction) of the total potential conflicts.

Pedestrian potential conflicts were largely consistent throughout the day. There were more pedestrians involved in potential conflicts during the weekend than during the weekday.

While the number of potential pedestrian-vehicle conflicts reduced, the average speed of vehicles making turning movements increased by approximately 4 mph. The pedestrian scramble removes pedestrians on the south leg crosswalk during the vehicle phase, and so the right- and left-turning vehicles are essentially free-flowing.

Overall, the analysis validated that there is a 75% performance improvement in safety between vehicles and pedestrians with the implementation of the pedestrian scramble. It also provided the safety engineers with insights into further modifications required to reduce vehicle speed.

Discussion/Deliverables: The group is now considering additional safety effectiveness studies at new intersections. The same safety analysis was implemented at a busy unsignalized intersection on Peachtree Street and a freeway on ramp intersection to identify and quantify aggressive driver behavior (8 mph right-turn speed), who arrives first (the pedestrian or the vehicle) and pedestrian high-risk conflicts to assist Aecom with justification for improvement recommendations to Midtown Atlanta Alliance.

Conclusions: Midtown Atlanta reduced the time to effectively validate pedestrian safety performance measurements at this new scramble intervention.

The project implementation team, led by Aecom, were very satisfied with the quick and effective way to evaluate and validate "before and after" pedestrian safety numbers at the intersection of Spring Street and 5th Street in midtown Atlanta where countermeasures have been deployed. The results were used to justify the intervention cost. Aecom recently co-presented with Brisk Synergies the results of the case study in a 40 min webinar.