Hardware-in-the-Loop Simulation Evaluation of Adaptive Signal Control


Dr. Seung Kook Wu, Georgia Tech  
Matthew Roe, MSCE, Georgia Tech
Dr. Michael Rodgers, Principal Research Scientist, Georgia Tech
Dr. Michael Hunter, Associate Professor, Georgia Tech

Project Overview:

Increasingly, adaptive signal control and other non-traditional solutions are being implemented in an attempt to improve signal system efficiencies, reduce congestion, enhance signal control responsiveness to incidents, and reduce signal re-timing costs.  As part of a recent adaptive signal control implementation in Cobb County, Georgia, a before-and-after operational comparison of an optimized time-of-day (T.O.D.) and an adaptive control system was undertaken.  The focus of this operational analysis was typical operating performance during the weekday peak, weekday off-peak, and weekend travel periods.  The initial study resulted in a general conclusion that under typical existing traffic conditions, both the well calibrated T.O.D. and adaptive control systems provided good, although similar, performance. Numerous others have also conducted field studies of various adaptive control systems.  A lack of consensus seems to be present regarding the benefits of adaptive control, with results often dependent on site specific issues, quality of before timing plans, selected performance metrics, etc.

However, in addition to operational performance under current traffic conditions the evaluation and selection of a signal control system should consider the system responsiveness to normal day-to-day variation in traffic demands, special events, incidents, future traffic demands, and other conditions.  A field data based study will typically not allow for an evaluation of control types under such conditions.  Non-recurring and future growth performance testing is more directly suited to simulation based evaluations.  Simulation is often utilized for the evaluation of Intelligent Transportation Systems (ITS) because it allows for controlled experiments to be conducted and statistically strong conclusions to be drawn, particularly in instances where it is not feasible or possible to use field data for an evaluation. However, potential drawbacks exist when utilizing simulation to evaluate cutting-edge or proprietary ITS implementations. First, no software emulation of the ITS in question may be available, and second, the software emulation of the ITS, if available, is not guaranteed to be an accurate representation of its real-world counterpart.

To overcome the limitations of field data or simulation only studies, hardware-in-the-loop simulation (HILS) may be utilized.  In traffic engineering the concept of HILS typically involves combining a microscopic traffic simulation program with a real-world traffic controller.  By enabling the controller to interact with the simulation model, the benefit of simulation analysis is gained while retaining the full functionality of the real-world system.  There have been numerous successful applications of HILS in the development, testing and evaluation of ITS in the transportation environment.

The schematic in Figure 1-a represents a notational field implementation of adaptive control and T.O.D. control.  The controllers residing in field cabinets communicate through fiber optic cables and modems.  The schematic shown in Figure 1-b is the adopted HILS system that replaces the physical roadway network with a simulated one.  The detector actuations from vehicles in the VISSIM model are sent out to the real-world controllers. Signal state information from the real-world controllers is accepted by the model.  Between the model and the real-world controller is a hardware device that facilitates the communication between the two, called a Controller Interface Device (CID).  

This research examined the initial findings of a HILS evaluation of ACTRA (i.e. T.O.D.) and SCATS (i.e. adaptive) signal control on an eleven-intersection section of Cobb Parkway, in Cobb County, Georgia.  It was seen that the HILS test bed generally provided a reasonable representation of both the T.O.D. and adaptive signal control performance.  However, it was noted that the adaptive control AM HILS results contain some unexpected behaviors, not observed in the field, in the southbound direction.  The explanation currently under investigation is that the responsive nature of adaptive control may be resulting in poor performance when the simulation contains a short term unrealistic behavior, whereas the relationships appear to indicate that for the system studied, the adaptive control may be more responsive to overall traffic demand. Initial indications are that during peak conditions, both control strategies provide similar performance, however, during the hours bordering the peaks, adaptive control is likely able to provide control more tailored to the current conditions.  Future efforts will further explore the findings.  In addition, the next phase of this effort will expand the scenario analysis to include growth scenarios, non-recurring congestion, and special events.

Sponsored by Cobb County Department of Transportation