One of the primary obstacles to improving the performance of signalized arterials has been the difficulty of gathering accurate and reliable data to assess arterial traffic conditions. The need to manually collect the data and then calculate performance metrics for individual intersections or arterials has made performance assessment a time-consuming and expensive process for transportation agencies.
Researchers in the University of Minnesota’s Department of Civil Engineering and Minnesota Traffic Observatory (MTO) developed a new system that automates data collection and performance assessment in real time. Because it can also refine the traffic signal parameters intelligently using archived data, the system has been dubbed “SMART-Signal,” the acronym for Systematic Monitoring of Arterial Road Traffic Signals. The effort was funded by the ITS Institute, Minnesota Local Road Research Board, and Minnesota Department of Transportation, with significant in-kind support from Hennepin County.
Principal investigator Henry Liu’s research is focused on traffic control and operations, network modeling, and simulation of transportation systems. Since joining the University of Minnesota faculty in 2005, Liu has led a series of research projects focusing on various components of the SMART-Signal system. Chen-Fu Liao, a senior systems engineer at the MTO, has also participated in the projects.
The SMART-Signal system is intended to be installed at a series of intersections along an arterial road. A dedicated microprocessor module is installed in the signal control cabinet at each intersection, interfacing directly with the cabinet without interfering with signal operations. SMART-Signal collects two types of event data: signal-phase change events and vehicle-detector actuation events. Event data are then packaged and transmitted in real time to the server located at the MTO.
Queue length is inarguably the most important performance measure at a signalized intersection. A major shortcoming of traditional input-output models used to estimate queue length has been their inability to determine queue length under saturated conditions—i.e., when the queue of cars waiting to pass through an intersection extends beyond the upstream vehicle detector. Under saturated conditions, data on incoming traffic flow are no longer available and the input side of the input-output model breaks down.
The SMART-Signal developers overcame this limitation by developing a new algorithmic approach to queue length estimation based on the mathematical properties of shockwaves. A shockwave forms whenever the density of vehicles in a traffic flow changes, such as when traffic is forced to stop at an intersection. This queuing shockwave will continue to propagate upstream as more and more vehicles arrive at the end of the queue. When the signal phase changes to green, however, vehicles will begin to depart from the head of the queue, forming a discharging shockwave; because this wave propagates more quickly than the first under most traffic conditions, it will eventually meet the front of the first shockwave. At that point, the queue disappears, as all vehicles are again moving forward. If all the vehicles from the queue are not able to pass through the intersection before the start of the next red phase, a residual queue forms and begins to propagate upstream, starting the cycle again.
SMART-Signal uses data from the upstream vehicle detector to identify three critical points in the shockwave propagation: when the queuing shockwave reaches the upstream detector; when the discharging shockwave passes the detector; and, most important, when the rear of the queue of stopped vehicles passes the detector. Before the rear of the queue reaches the detector, vehicles are generally moving under saturated traffic conditions, so they are closely spaced and there is little variance in the size of inter-vehicle gaps. But as the queue dissipates, vehicle spacing increases and the size of inter-vehicle gaps becomes more variable. These characteristic changes make it possible for SMART-Signal to identify the moment when the shockwave corresponding to the rear of the vehicle queue passes the upstream detector.
Using these three critical points, SMART-Signal’s estimation algorithms can estimate traffic states upstream from the intersection and determine the maximum length of the vehicle queue, an important performance metric for arterial intersections. Extensions to the traffic model allow the system to work with multiple-lane traffic and with detectors that produce second-by-second data rather than event data.
For measuring the performance of a corridor or network, the key metric is travel time. Measuring travel time on signalized arterials is difficult because unlike the smooth traffic flow observed on highways under normal traffic conditions, traffic on arterials is repeatedly interrupted by signal changes.
SMART-Signal approaches this problem using the same sources of event-based data to simulate the movements of a virtual “probe vehicle” along the arterial road. Data from individual intersections are fed to a central processor to create an estimated model of the state of traffic along the entire route, including signal states and queue lengths. Once this model has been estimated in a simulation system, a virtual probe vehicle is created and allowed to move along the simulated route. As the virtual probe moves, it can modify its own state in response to the state of traffic around it by accelerating, decelerating, or maintaining a constant speed at each time step as it encounters queues, traffic signals, and changes in traffic density.
The virtual probe approach to travel time estimation actually benefits from traffic flow interruptions caused by traffic signals, because differences between the trajectories of the virtual probe vehicle and a hypothetical real vehicle—for example, if the virtual probe moves slightly faster than real-world traffic—are corrected by stopping at a red traffic signal. In this way, even an error that has increased in size as the virtual probe passed through several intersections will be eliminated at a single blocked intersection. This error-correction property increases the robustness of the travel-time estimation model and makes the result less sensitive to model parameters.
Because of the variability of driver behavior in the real world, the travel time results of this estimation model are best expressed as an estimated interval of travel times.
SMART-Signal was first implemented on two arterial corridors in the suburban Twin Cities metropolitan area: at 11 intersections on France Avenue in Bloomington and Edina and six intersections on Trunk Highway 55 in Golden Valley, where the system’s ability to estimate queue lengths and arterial travel times was validated under real-world conditions. The system has also been deployed at three intersections on Prairie Center Drive in Eden Prairie, Minn.
The success of the initial SMART-Signal deployment garnered the 2009 CTS Research Partnership Award for the researchers and agency personnel involved in the development and implementation of the system. The University of Minnesota also filed a patent to protect the intellectual property on the SMART-Signal system in 2009. Recently, the University has issued a temporary license to Iteris Inc. in California for a demonstration project in Pasadena.
In the current phase of research, Liu and his colleagues are focusing on the development of a new version of SMART-Signal hardware that will be easier to deploy and maintain. The prototype system was not optimized for use by agency personnel. The new version is designed to integrate more easily with the new signal controller cabinets being used by Mn/DOT, and it also includes a graphical user interface for installation and maintenance. The next-generation SMART-Signal system will be installed and tested on two suburban trunk highway corridors.