Principal Investigator

Bernardo Pires

Status

Completed

Start Date

Jan. 1, 2016

End Date

Dec. 30, 2017

Project Type

Research Advanced

Grant Program

MAP-21 TSET National (2013 - 2018)

Grant Cycle

2016 TSET UTC

Visibility

Public

Abstract

Surtac, the real-time adaptive traffic signal control system, has been demonstrated to significantly improve traffic flow on multiple performance metrics, including reductions of 25% of travel time and 40% wait time for motor vehicles. The objective of this project is to bring this same intelligence to pedestrian traffic, which has, thus far, not been targeted by Surtrac deployments. Phase 1 of this two-year project will analyze pedestrian traffic at multiple Surtrac deployments. Phase 2 will focus on an intersection already equipped with Surtrac system in the Oakland / East Liberty region and will add additional sensing and processing capabilities to determine the presence of pedestrians waiting to cross the intersection. Phase 3 will focus on the expected Surtrac deployment in Pittsburgh’s downtown and will focus on pedestrian density as a more fine grained input for the Surtrac scheduler.    

Description

Surtrac (Scalable URban TRAffic Control) is a real-time adaptive traffic signal control system developed in the Intelligent Coordination and Logistics Laboratory at Carnegie Mellon University (CMU) as part of CMU’s Traffic21 Research Initiative.  Surtrac combines new concepts from the fields of artificial intelligence and traffic theory, and is designed specifically for optimizing traffic flows in urban road networks, where there are multiple, competing dominant flows that shift dynamically through the day. In contrast to most commercial adaptive traffic control systems, Surtrac takes a totally decentralized approach to traffic control. Each intersection allocates its green time independently based on actual incoming vehicle flows, as seen through video or radar detection, and projected outflows are then communicated to neighboring intersections to increase their visibility of future incoming traffic. Reliance on decentralized intersection control ensures maximum real-time responsiveness to actual traffic conditions, while communication of projected outflows to neighbors enables coordinated activity and creation of green corridors. The system is inherently scalable to road networks of arbitrary size, since there is no centralized computational bottleneck. 

An initial pilot implementation and field test of Surtrac was carried out in June 2012, on a nine-intersection road network in the heart of the East Liberty region of the City of Pittsburgh where Penn Avenue, Centre Avenue and Highland Avenue intersect. A series of “before” and “after” drive through runs were performed at 4 different periods of the day, and various performance metrics (travel time, speed, number of stops, wait time, fuel consumption, emissions) were computed for each test condition. Across all metrics studied, Surtrac was shown to produce significant performance improvement, including an overall 25% reduction in travel times, a 40% reduction in wait times, and a projected reduction in emissions of over 20%. In November 2013, the pilot deployment was expanded to include nine additional intersections moving eastward along Penn Avenue to the major cross-corridor at Fifth Avenue. Near identical additional improvement was observed when an analogous evaluation was performed. In early 2015, the deployment was further expanded to include 6 additional intersections moving east on Penn Avenue to Braddock Avenue at the City’s border. In September 2015, the size of the Pittsburgh Surtrac deployment was doubled, this time adding 23 intersections moving west from the pilot deployment site along the Centre Avenue and Baum Boulevard corridors to Craig Street in Oakland. A performance analysis of this last expansion is currently underway, and plans are in place to add 2 additional intersections by the end of the year, which will bring the size of the deployment to 49 intersections. See Figure 1 on Supplementary Material for the Surtrac pilot sites and Figure 2 for the Surtrac expanded deployment. The Surtrac work has been funded by The Hillman Foundation, The Heinz Endowments, The Richard King Mellon Foundation, UPMC, the CMU T-SET University Transportation Center and CMU The Robotics Institute. 

Surtrac’s current system is tailored to optimize motor vehicle traffic, as this is the critical component to enhance flow, reduce pollution, and improve air quality. Nonetheless, to continue the expansion of Surtrac into areas with heavier pedestrian traffic, such as downtown Pittsburgh, it becomes critical to make Surtrac “aware” of pedestrians. To this end, it is necessary to equip Surtrac with both new hardware, in the form of video cameras directed at sidewalks, and software, in the form of novel algorithms for detection of presence and density of pedestrians. 

We expect our industry partner, Axis Communications, to provide us with the necessary hardware cameras (although it is not clear yet if this partner will be able to provide camera installation and setup). The software shall be developed by this project and will leverage the prior PI experience on the T-SET sponsored project “Automatic Counting of Pedestrians and Cyclists”, which has deployed a hardware prototype (see Figure 3) and developed state-of-the-art pedestrian detection software, achieving approximately 95% accuracy on a published dataset. A second T-SET sponsored project, “Measuring Pedestrian Wait-Time at Intersections”, as used the same hardware device to collect data that closely resembles the expected point of view for this project (see Figure 4), and will provide guidance as to how best deploy the algorithms in a robust and reliable way. 

Phase 1 of the proposed project will run during the first five months of 2016. During this Phase, a detailed analysis of the currently deployed Surtrac intersections will be conducted, including estimating the pedestrian flows in at least 10 locations. These locations will include Centre/Negley, Baum/Negley, and Baum/Roup, which all have significant pedestrian traffic. The Centre/Negley intersection probably has the highest pedestrian traffic of all Surtrac locations. It has a separate “all ped” phase which means that only pedestrians can cross (and they can cross in any direction during this phase). The Baum/Negley intersection is next intersection going north on Negley. Here the pedestrians travel as part of the associated vehicle phase (as is normally the case). Finally, Baum/Roup is a more minor intersection that forms a triangle of sorts with the above two. Since Baum is somewhat of a throughway, both this intersection and Baum/Negley must deal with pedestrian cross traffic. 

Phase 2 of the proposed project will run from June 2016 to March 2017. This phase will deploy up to 8 video cameras and develop a computer vision system to detect the presence of pedestrians in one of the previously analyzed intersections. During this Phase, an API will be developed for interfacing between the Surtrac system and the Computer Vision system. 

Phase 3 of the proposed project will run from April to December of the 2017 calendar year and will focus on a deployment in the Downtown area. Although the specific site is to be determined, it will surely feature heavy pedestrian traffic. During this Phase the Computer Vision algorithms will be improved to not only detect the presence of pedestrians, but also the density of pedestrian traffic. Such information will provide a more fine-grained input for Surtrac, and will affect directly the Surtrac core scheduler.

Timeline

Phase 1 – January to May 2016
- January 1, 2016 to January 31, 2016: Selection of 10 target intersections. ? February 1, 2016 to March 31, 2016: Data collection on target intersections. ? April 1, 2016 to May 14, 2016: Analysis of data collected. ? May 15, 2016 to May 31, 2016: Selection of Phase 2 deployment site

Phase 2 – June 2016 to March 2017
- June 1, 2016 to August 31, 2016: Deployment of cameras and additional hardware at selected
deployment site in Oakland / East Liberty area
- July 1, 2016 to December 31, 2016: Development of computer vision software for detection of pedestrians
- October 1, 2016 to March 31, 2017: Development of Surtrac API for interfacing with pedestrian presence detection system.

Phase 3 – April 2017 to December 2017
-  April 1, 2017 to June 30, 2017: Deployment of cameras and additional hardware at selected
deployment site in Pittsburgh’s Downtown
- July 1, 2017 to December 30, 2016: Development of computer vision software for estimation of pedestrian density at the intersection
- July 1, 2016 to December 31, 2017: Extend API (and core Surtrac scheduler) to accept and deal with pedestrian density information.

Strategic Description / RD&T

Deployment Plan

See Project Timeline. Deployments are planned during Phase 2 (Oakland / East Liberty) and Phase 3 (Downtown)

Expected Outcomes/Impacts

- Detailed analysis of 10 current Surtrac deployments 
- Deployment of pedestrian detection at two Surtrac intersections 
- Development of Computer Vision algorithms for the detection of presence of pedestrians and for
the estimation of the density of pedestrians
- Integration with Surtrac API for pedestrian presence and extension of same API and core scheduler to handle pedestrian density information

Expected Outputs

TRID

Individuals Involved

Email	Name	Affiliation	Role	Position
jgb@rapidflowtech.com	Barlow, Greg	SCS	Co-PI	Faculty - Research/Systems
kocamaz@cmu.edu	Kocamaz, Mehmet	SCS	Co-PI	Faculty - Research/Systems
bpires@cmu.edu	Pires, Bernardo	SCS	PI	Faculty - Research/Systems
sfs@cs.cmu.edu	Smith, Stephen	SCS	Co-PI	Faculty - Research/Systems

Budget

Amount of UTC Funds Awarded

$84743.00

Total Project Budget (from all funding sources)

$84743.00

Documents

Type	Name	Uploaded
Final Report	37_-_Pires_PedDetectionForSurtrac_FinalReport.docx	June 19, 2018, 4:32 a.m.

Match Sources

No match sources!

Partners

No partners!