Results of a real-time, decentralized traffic signal system pilot test in Pittsburgh, PA.
Pittsburgh, Pennsylvania, United States
Real-Time Adaptive Traffic Signal Control for Urban Road Networks: The East Liberty Pilot Test
Summary Information
The Robotics Institute at Carnegie Mellon University (CMU) developed an adaptive traffic signal control system, SURTRAC (Scalable URban TRAffic Control), designed specifically for urban road networks, where there are multiple, competing dominant flows that shift dramatically based on time of day. This system is a decentralized system, where each signal independently determines its phasing based on incoming vehicle flows and projected outflows from neighboring signals.
In order to test this system, a pilot of nine intersections in the East Liberty area of Pittsburgh, Pennsylvania was equipped with this technology for a before and after study to analyze the difference in performance for the existing and SURTRAC systems. The test area has an average daily traffic volume of nearly 30,000 vehicles per weekday.
Methodology
To evaluate the SURTRAC system's performance, a series of timed, drive-through runs of the pilot test were conducted both under existing conditions and with the SURTRAC system enabled. The twelve highest volume routes through the East Liberty pilot test site were chosen for analysis. Tests were conducted for four weekday time periods -- AM-Rush, Midday, PM-Rush, and Evening -- in order to test the performance of the system in multiple conditions. "Before" testing was conducted in March 2012 and "after" testing was conducted in June 2012, with slightly higher (roughly 5%) traffic volumes in June.
When the SURTRAC runs are compared with existing conditions, the level of improvement during all time periods was significant. A summary of average improvements by time period can be found in Table 1 below. Travel times were reduced 17 to 33 percent. Vehicle speeds increased 27 to 49 percent. The number of stops was reduced by 9 to 53 percent, with improvements particularly significant during off-peak periods. Wait times were reduced by 28 to 50 percent on routes within the pilot area. The midday period saw the largest improvements across all categories, demonstrating the performance improvements of SURTRAC over actuated free mode during the midday travel period.
% Improvement | Travel Time | Vehicle Speed | Number of Stops | Wait Time | Emissions/Fuel Consumption |
AM Rush | 30.11% | 33.78% | 29.14% | 47.78% | 23.83% |
Mid Day | 32.83% | 48.55% | 52.58% | 49.82% | 29.00% |
PM Rush | 22.65% | 27.45% | 8.89% | 35.60% | 18.41% |
Evening | 17.52% | 27.81% | 34.97% | 27.56% | 14.01% |
Overall | 25.79% | 34.02% | 31.34% | 40.64% | 21.48% |
While not every route saw improvement for all five variables during all time periods, no route had its performance adversely affected during all four time periods. The routes that saw reduced performance under SURTRAC operations had generally previously been the emphasized movements under the fixed signal timing plans for the peak periods. The results by time period can be found in Tables 2-5 in the report.