This paper develops a distributed-coordinated methodology for traffic signal timing optimization problem. Our formulation and solution methodology distribute the network level signal timing optimization problem to intersection level. As a result of this distribution, the complexity of the problem is significantly reduced so that it becomes real-time and scalable. We formulated a mathematical programing model for each intersection, based on the cell transmission model and created coordination between them to avoid finding locally optimal solutions. The neighboring intersections coordinate their decisions to avoid long queues. The proposed formulation accounts for oversaturated conditions with using an appropriate objective function and explicit constraints on queue length. We also proposed a rolling horizon solution algorithm and applied it to several case study networks under various demand patterns. The algorithm controlled queue length and maximized intersection throughput. We compared the solution to solutions provided by a central optimization approach and observed differences of less than 1% on a number of performance measures over different conditions.

This paper studies intersection signal control in which traffic arrivals from all approaches along with the queueing situation are assumed known. The optimal control policy minimizes the overall intersection delay by deciding the green intervals for signal phases dynamically as driven by real-time traffic but subject to a set of constraints. The set of constraints allow consideration of practical factors such as min/max green time for each phase and may be flexibly pre-defined. This paper proposes an analytical model for intersection vehicle delay and derives optimal conditions for green signal switch. Two numerical algorithms are proposed: optimum based (DORAS) and queue-based heuristic respectively. Numerical tests are conducted via simulation using an actual intersection data covering peak hours, mid-day hours and night hours. With a fixed sequence of phases, the tests show an up to over 20 percent reduction of average vehicle delay due to our proposed algorithm compared with the fully actuated control. This research prepares a new concept for the signal control problem in a cyber-enabled environment.

Various Adaptive Traffic Control Systems (ATCSs) have been evaluated in the past. Researchers have usually shown either positive or neutral impacts of these dynamic signal strategies. In most cases, such studies were performed in the field. Nevertheless, due to the high costs for performing the field-based experiments, many field ATCSs evaluations were performed in anecdotal way. While simulation models often provide a variety of outputs, they are frequently questioned for their ability to accurately represent field conditions. This study aims to address these issues by presenting a novel approach for developing a high fidelity microsimulation model for an urban network, using different types of the field data for calibration and validation purposes. In order to provide further evidence that the presented model is an accurate virtual representative of field conditions, an additional validation procedure is conducted through the comparison of modeled and observed durations of average green times. Average green times collected from the field are not based on green times programmed in the field controllers, but based on the actual field green times collected from the controllers’ logs. The calibrated and validated simulation model is further used to evaluate the performance of Kadence ATCS for the first time in a microsimulation.

Real-time traffic management with environmental objectives has been a difficult challenge due to (1) the highly dynamic and uncertain nature of road traffic and their emission profile; (2) the need for generating timely and robust decisions for large-scale networks; (3) the incorporation of potentially massive and heterogeneous on-line data; and (4) the balance between traffic and environmental objectives. To address these challenges, this paper proposes a real-time traffic signal control framework based on linear decision rule (LDR), which is integrated with realistic traffic and emission modelling via microsimulation. The implementation of the LDR method is divided into an off-line module and an on-line module. The off-line module trains the LDR using historical traffic data, which amounts to a data-driven distributionally robust optimization with traffic and environmental objectives. The on-line module efficiently implements the trained LDR, whose performance is guaranteed by the off-line training. Our study has demonstrated the capability of the linear decision rule approach to (1) explicitly and accurately account for the traffic and environmental objectives in on-line operation; and (2) integrate with realistic traffic microsimulation to perform real-time signal control on potentially large-scale networks.

Recently, strategies of decentralized logic have been developed to tackle the problem of traffic congestion in urban networks. Such strategies approach the network-wide control problem through local (by junction) actions, but are aimed to improve traffic flow efficiency at network level, with low design effort and infrastructure investment. This paper presents, compares, and evaluates two such innovative approaches proposed in literature. The first, a job scheduling algorithm, comprises the basis of the SURTRAC (Scalable URban TRAffic Control) system, while the second is the known max or back pressure algorithm. The paper compares also their performance against TUC (Traffic-responsive Urban Control), a well-established strategy of centralized logic, developed to provide coordinated control in large-scale networks. For evaluation purposes, the AIMSUN simulation model of the city center of Chania, Greece, is used. The results of the study indicate that only TUC and max pressure retain their performance independent of the prevailing traffic conditions, being also computationally simpler than job scheduling. As far as the sensing infrastructure is concerned, both decentralized approaches are demanding, as they require frequent and accurate measurements. TUC is less demanding in this respect, but calls for communication lines between the junction controllers and the central computer. Last not least, compared to both decentralized approaches, TUC provides a sequence of signal plans with less excessive differences among each other, thus fewer disturbances to the common network users. Nevertheless, more investigations, including field trials, would be needed for more comprehensive conclusions concerning the strong and weak elements of each approach.

Congestion at the downstream of a freeway off-ramp often propagates the traffic queue to the mainline, and thus reduces the freeway capacity at the interchange corridors. To prevent such potential queue spillover and improve the traffic control efficiency over the entire corridor, this study develops an integrated control system which includes three primary functions such as off-ramp queue estimation, arterial adaptive signal control, and freeway off-ramp priority control. Based on detected flow data, the system firstly estimates the queue length on the target off-ramp. If no potential queue spillover is predicted, the adaptive signal control function will adjust intersection signal timings and provide dynamic signal progression to critical path-flows. Otherwise, the off-ramp priority control function will be activated to clear the queuing vehicles on the off-ramp. To evaluate the effectiveness of the proposed system, this study conducts numerical studies on a field interchanged corridor with a well-calibrated simulation platform. The experimental results reveal that the overall network performance is improved by the proposed control system, compared with other existing systems. Also, the examination of freeway time-dependent travel time distribution proves the effectiveness of the proposed system in preventing off-ramp queue spillover.

In this paper, a game based adaptive signal control-Active-Responsive Signal Control-is proposed. The design enables active cooperation from the connected vehicles and thus is able to take full advantage of the emerging Connected Vehicle (CV) technology. The signal control problem is modeled as a Stackelberg game which is then further formulated as a bi-level optimization problem. The upper level optimizes the game leader’s (signal) decision on signal timing plan while the lower level optimizes the game followers’ (connected vehicles) decisions on route choice. The inner-outer iterative evolutionary method is applied to solve the bi-level optimization. The proposed signal control requires no automation from the connected vehicles and is beneficial as long as connected vehicles market penetration is above zero. Hence, it is with great potential during the long-term roll-out of the Connected Vehicles technology. The benefit of the proposed system was evaluated at an isolated four-leg eight-phase hypothetical intersection. The evaluation results showed that the proposed system outperforms the traditional traffic signal control method with passive connected vehicles in terms of total intersection delay and average vehicle delay.  The reduction in the average vehicle delay ranges from 8% to 14% under different CV market penetrations.

Utilizing vehicle infrastructure integration (VII) data to characterize the dynamic traffic system is an emerging research topic, given the development of VII technology. Dynamic traffic flow prediction is one of the most important aspects of adaptive control algorithms or systems. Prediction results directly affect the control effect of adaptive signal control systems. A traditional traffic flow prediction model is limited to fixed detectors. This model can identify vehicles but not the status of vehicles. The location, identification number (ID), and speed of a vehicle are easily obtained in VII environments. Thus, a dynamic traffic flow prediction model was proposed based on VII. This work investigates the factors affecting model accuracy, including the location of upstream VII equipment, road segment length, traffic volume, and tuning directions. The effect of these factors on model performance is demonstrated through a simulation analysis. Analysis results show that these factors variably affect the proposed model, although the mean relative error of prediction is less than 10% with various factors. The proposed model, which was adopted to an adaptive signal control algorithm, and the actuated control were compared. Simulation results indicate that adaptive signal control based on the proposed model outperformed the actuated control by reducing average delay by as much as 19.8%. The reliability of the proposed model was verified, and the model can be applied in adaptive control algorithms or systems.

The Florida Department of Transportation (FDOT) has proposed the implementation of Adaptive Signal Control Technology (ASCT) in Florida to overcome the limitations of traditional signal sytems in cases of changes in traffic demand, weather, incidents, etc.

The performance of InSync and SynchroGreen along four corridors in Florida is first evaluated. Field data are used to build regression models of performance measures as functions of site characteristics. Meta-analysis is then conducted using data from a similar study by Virginia Department of Transportation (VDOT). Cluster analysis is used to find and compare results of sites with similar characteristics. Additionally, the regression model developed from Florida data is used to predict the travel time change for Virginia. Both results support the validity and spatial tranferability of the developed relationship between site characteristics and ASCT benefits.

Selecting an appropriate adaptive controller in optimizing isolated traffic signals in urban congested road networks with transit priority and multiple incidents is a challenging task. This is because the decision is likely to affect transit trips, overall network throughputs, and efficiencies in different ways. This paper presents a new purely distributed control logic based on a person-based hypothesis aiming to enhance area-wide control without any centralized infrastructure. The control logic model was applied to a large network of 49 intersections under various demand patterns in a simulation environment. The performance of the control logic was compared with that of the existing control system with five phase settings: dual actuated, protected actuated, protected pre-timed, split actuated, and split pre-timed. Except for an extremely congested demand case, the proposed control logic produces better transit bus trips while having comparable performance in efficiencies compared with existing systems. The results of the sensitivity analysis indicate very high potential of using the unexplored capacity of the system with increased bus trips and lesser delays.

ACS-Lite is a member of the adaptive signal control family featuring a cost-effective, widely deployable system for linear or arterial network traffic management.  There are a number of parameters that a user must configure for an ACS-Lite deployment and few studies have been conducted to provide guidance on the configuration of those parameters.  Likewise, when retrofitting an existing signalized corridor with ACS-Lite, the existing vehicle detection may not conform with the recommended setup, which can be costly to change.  This study evaluated one parameter configuration and two vehicle detection configurations in a simulation environment based on a congested corridor of five intersections in Morgantown, WV.  The evaluation was based on the analysis of percent arrivals on green and red block delay, derived from high resolution traffic signal data.  The findings indicate that it is preferable to use the most relaxed offset increment (6 seconds) and maximum deviation (unbounded) settings to allow the algorithm to adjust.  Under these settings, the algorithm was able to remedy a poor intersection offset.  The need for lane-by-lane stop bar detection for phase utilization did not seem critical compared to approach-based detection, assuming the multi-lane phases are predominantly on the mainline movement.  The algorithm performance seemed to suffer if vehicles queue over the upstream flow profile detectors, so it is recommended that they be located upstream far enough to avoid queueing. 

Adaptive traffic control systems vary in their requirements for data to drive their logic, however it is common for a system to use vehicle counts and occupancy values as inputs to their control algorithms, in addition to vehicle presence.  Historically, pavement loop detectors have been used for providing data to adaptive systems. Due to the costly installment, involved lane closures and maintenance difficulties of loop detectors, many agencies would like to replace them with video detectors, however the performance of video detection with adaptive control has not been trouble free.  Most of the existing comparative studies assess the performance of video detectors at a microscopic level, where individual actuations (on and off times) are compared against ground truth data. There is scarcity of research that evaluates video detectors at a macroscopic level on metrics such as counts, occupancies, and time to gap, even though adaptive traffic control systems use these aggregated measures as detection inputs for their control algorithms. This study fills that gap by evaluating the performance of stopbar mounted advance video detectors at a macroscopic level, providing discussion and recommendations on using these detection devices for adaptive control inputs.

This research presents a signal control strategy utilizing vehicular trajectory-driven optimization method. Band Finder provides individual drivers equipped with opt-in device (i.e. smartphone or tablet) with advisory speed information enabling to minimize overall travel time of the vehicle while negotiating signalized corridor. Signal status parameters such as cycle length and remaining green/red time are continuously captured. At the same time in-vehicle unit provides vehicle position information through cell-phone GPS receiver. Both inputs are then used by optimization algorithm to provide optimal vehicle speed that will achieve minimal vehicle delay along the signalized corridor. To calculate the most optimal advisory speed for individual vehicles a non-linear optimization algorithm was utilized. The concept was evaluated using microsimulation in PTV VISSIM. The results for selected signalized corridor in Woodbridge, New Jersey indicate 1.2% to 5.4 % reduction in overall corridor travel time depending on different market penetration and traffic volume conditions.