2021

Weimin Jin (Weimin.Jin@arcadis.com), Arcadis US Inc
M. Sabbir Salek, National Center for Transportation Cybersecurity and Resiliency (TraCR)
Mashrur Chowdhury, Clemson University
Mohammad Torkjazi, University of South Carolina
Nathan Huynh, University of Nebraska, Lincoln
Patrick Gerard, Clemson University

Adaptive Signal Control System (ASCS) is an advanced traffic signal control system, which can adjust signal timings based on real-time and/or predicted traffic demand. The operational benefit of ASCS may vary depending on types of ASCS, corridor characteristics, and geographical areas. This paper evaluates the operational performance of 11 ASCS corridors located throughout South Carolina (SC). These corridors are operated using SynchroGreen, one of several types of ASCS, developed by TrafficWare. Based on the operational analysis, it is found that when ASCS is operational, it reduces the travel time on the corridor by an average of 6% and improves travel time reliability by an average of 26% compared to when the non-ASCS traffic signal control system (e.g., pre-timed and actuated signal control) is operational. ASCS reduced travel time on a corridor by 61% on average during the day and by 77% on average during peak periods. Additionally, ASCS improved travel time reliability by 53% on average during the day and by 52% on average during peak periods. The operational effectiveness of ASCS in reducing travel time and improving travel time reliability is consistent in both directions for eight corridors and five corridors, respectively. Lastly, ASCS is found to produce higher operational benefits in terms of reducing travel time if the average speed of an ASCS corridor is lower than or equal to 35 mph and the number of signals on the ASCS corridor is more than ten.

Andalib Shams (ashams@iastate.edu), Iowa State University
Christopher Day, Iowa State University

Advanced signal control algorithms are anticipated with the increasing availability of vehicle speed and position data from vehicle-to-infrastructure communication and from sensors. This study examines the impact of the sensing range, meaning the distance from the intersection that such data can be obtained, on the quality of the signal control. Two signal control methods, a Self-Organizing Algorithm (SOA) and Phase Allocation Algorithm (PAA), were implemented in simulation and tested to understand the impact of sensing ranges. SOA is based on fully-actuated control with secondary extension for vehicle platoons along the arterial. PAA uses dynamic programming to optimize phase sequences and duration within a planning horizon. Three different traffic scenarios were developed: symmetric, asymmetric, and balanced.  In general, both algorithms exhibited improvements in performance as the sensing range increased. Under the symmetric volume scenario, SOA converged at 1000 ft and PAA converged at 1500 ft. In asymmetrical and balanced scenarios, both algorithms outperformed conventional methods. Both algorithms performed better than coordinated-actuated control if the sensing range is 660 ft or higher. For low sensing ranges, SOA experiences similar delay compared to fully-actuated control with advance detector and PAA experienced more delay than coordinated-actuated control in symmetric and asymmetric scenario but performed better for asymmetric scenario. For the SOA, the sensing range may constrain the maximum allowable secondary extension, hence as the sensing range increases, the vehicular delay would decrease for arterial but increase for non-arterial movements. For PAA, arterial and non-arterial delay decreases with increase in sensing range until it converges.

SMA Bin Al Islam, North Carolina State University
Mehrdad Tajalli, United Parcel Service (UPS)
Rasool Mohebifard, North Carolina State University
Ali Hajbabaie (ahajbab@ncsu.edu), North Carolina State University

Sharing real-time data between connected vehicles (CV) and traffic control infrastructure provides an opportunity to improve the performance of adaptive signal control strategies. However, the effectiveness of such strategies would depend on the level of traffic observability in a transportation network, which is a function of the CV market share and the presence of other means of detection. This paper aims to investigate the effects of various degrees of traffic observability on traffic operations, signal coordination, and travel time reliability in arterial streets with adaptive signal control. Specifically, we have incorporated loop detector and CV data into an adaptive signal control system and have measured several mobility- and event-based performance metrics under different degrees of traffic observability with various CV market penetration rates. A real-world arterial street of ten intersections in Seattle, WA is simulated in Vissim under peak hour traffic demand level. The results showed that a 40% CV market share was required for the adaptive signal control system based on only CV data to outperform a system that used only loop detector data. At the same market penetration rate, signal control with CV-only data resulted in the same traffic performance, progression quality, and travel time reliability that signal control with CV & detector data provided. The inclusion of loop detector data did not further improve traffic operations when the CV market share reached 40%. Integrating 10% of CV data with loop detector data in the adaptive signal system improved traffic performance and travel time reliability.

Zhensong Wei, University of California, Riverside
Peng Hao, University of California, Riverside
Kanok Boriboonsomsin, University of California, Riverside
Matthew Barth, University of California, Riverside

The eco-approach and departure (EAD) system for signalized intersections aims to provide speed recommendations for drivers or automated vehicles so that they can pass through intersections in an eco-friendly manner. However, most existing applications were developed for fixed-time signals, which was not adaptive to the massively deployed actuated signals in the U.S. In this paper, we developed an adaptive EAD strategy for human drivers or automated vehicle controllers to minimize the expected energy consumption when passing an actuated signalized intersection. The historical signal phase and timing (SPaT) data are applied to calculate the probability that one signal state (including phase status, time in the phase, minimum and maximum time-to-change) transfers to another state. A graph-based model is created with nodes representing dynamic states of the host vehicle (distance to the intersection and current speed) and signal state (passing time and estimated minimum and maximum time-to-change), and directed edges with weight representing expected energy consumption between two connected states. Then a dynamic programming approach is applied to identify the optimal speed for each vehicle-signal state iteratively from downstream to the upstream. Real-world SPaT data collected from the intersection of Wilmington Avenue and E Carson Street in Carson, CA is applied in the simulation, which has shown that the proposed method is robust and adaptive to varying traffic conditions, and achieves 40% energy savings when the vehicle arrives in the red time and 8.5% energy savings when the vehicle arrives in the green time compared to other baseline methods.

Peirong (Slade) Wang, University of Texas, Arlington
Pengfei (Taylor) Li (Taylor.Li@uta.edu), University of Texas, Arlington
Farzana Chowdhury, University of Texas, Arlington

In this paper, we present a novel adaptive traffic control strategy for urban signalized interchanges. The signalized interchanges refer to those controlled by a single traffic signal controller, but vehicles must cross two or more stop lines to clear. Examples include the diamond Interchange (DI), diverge diamond interchange (DDI), and single point urban interchange (SPUI). With the expansion of urban areas, such interchanges are increasingly common and often become mobility bottlenecks. The underlying theory for traffic signal optimization is the cumulative vehicle counting curves which assume that vehicles are no longer restricted once they cross the stop line. However, at a signalized interchange, vehicles stop multiple times before crossing. This phenomenon cannot be effectively reflected with the classic cumulative vehicle counting curves. The phasing sequence is also challenging due to the limited space within the interchange. Lack of supporting theories also makes it difficult to develop more advanced adaptive traffic control strategies for urban signalized interchanges. We propose an adaptive traffic control framework based on a novel traffic control representation, phase-time network . The objective of this system is to dynamically fine-tune control splits and optimize the phasing sequence based on real-time travel demands. The optimization problem is formulated into a mixed-integer linear programming (MILP) formulation. Then an efficient A-D curves-based algorithm is presented and evaluated within a hardware-in-the-loop traffic simulation environment. The proposed MILP formulation and algorithm are evaluated in numerical experiments. The results of all numerical experiments validate the formulation and show promise for real-world implementations.

Zicheng Su, City University of Hong Kong
Andy Chow (andychow@cityu.edu.hk), City University of Hong Kong
Renxin Zhong, Sun Yat-Sen University

This paper presents an adaptive traffic controller built upon an integrated model-based and data driven approach. Adaptive traffic controller adjusts control policies according to real time traffic variations. However, its practical effectiveness can be hindered by the computationally complexity associated with applications to large road networks influenced by uncertainties. In our proposed framework, the model-based component operates based on short-term state predictions derived from an underlying kinematic wave traffic model within a finite time horizon. The data-driven component is an approximate dynamic programming (ADP) which approximates future state and control interactions through iterative learning with actual realisations of traffic states. To address the computational issues, we further present a decentralised solution approach for calculating the network-side control policies over different locations in an asynchronous way. The proposed control algorithms are tested over networks with different topologies and scenario settings. It is first shown that the proposed controller is able to minimise both average delays and their variability in stochastic traffic networks with use of flexible acyclic timing plans. With incorporation of the data-driven ADP component, the results further show that the decentralised solution approach can deliver similar performance compared with its centralised counterpart in congested conditions with significantly reduced computational time. This study contributes to the design of adaptive network traffic control systems with advanced optimisation techniques.

Ziang He (230189675@seu.edu.cn), Southeast University
Pan Liu, Southeast University
Yu Han, Southeast University

The primary objective of this study is to exploit a hierarchical perimeter feedback control approach considering the queue management and density spread optimization and total order inflow distribution optimization based on the conception of macroscopic fundamental diagram (MFD). In literature, network perimeter feedback control approaches are proven to effectively improve traffic efficiency in real time with simple and easily accessible parameters. However, the distribution problem of total order inflow identified by the network control system has been neglected in previous MFD based perimeter control works. But the network density spread problem is important to further improve system mobility. To fill the gap, a novel network total order inflow distribution optimization algorithm considering queue management were exploited. Using the taxi GPS data and field survey data, a microsimulation simulation model was modelled to test the proposed optimization control approaches. Three scenarios including original scenario, sample control scenario (sample scenario) and distribution optimization control scenario (distribution scenario) were compared based on the microscopic simulation model in this work. The proposed distribution control scheme was demonstrated to achieve a better suppression of congestion through better perimeter inflow distribution, thus improving average delay per vehicle (upgrade 6.18 to the sample scenario) and average speed (upgrade 6.93% to the sample scenario) with preventing the occurrence of the queue spillback in perimeter intersection upstream links.       Keywords: Macroscopic Fundamental Diagram, perimeter control, network inflow distribution, feedback control algorithm

Wan Li (liw2@ornl.gov), Oak Ridge National Laboratory
Zhenhong Lin, South China University of Technology

Traffic signal control and cooperation are extremely important to alleviate traffic congestion in a large traffic network. This study develops a higher-order conflict graph approach for network-wide traffic signal cooperation. A conflict graph is applied to model the traffic signal configurations, which identifies the conflict and unconflicted movements for each intersection. In conflict graph, the node represents each movement and the weight of each node can be defined as traffic volume, queue length, fuel consumption, or any weighted combinations of these measurements. The calculation of the optimal green light duration and green light sequence (for different movements) is equivalent to find the maximum weight independent set (MWIS) in the conflict graph sequentially and assign optimal green time to that set. The conflict graph also provides a uniform and efficient way to connect traffic signal operations among nearby intersections spatially. We introduced the concept of the k-th order neighborhood to model the degree of connectivity between each movement to the movements at upstream or downstream intersections. Then, the weight of each node in the conflict graph is not only represent its own congestion level, but also relates to the traffic conditions of nearby intersections. By this way, the cooperation of multiple intersections can be realized by incorporating their spatial connectivity into conflict graph and solving the MWIS problem. A simulation network is built in SUMO to test the effectiveness of the proposed method. Results suggested that the proposed model outperformed other state-of-the-art signal control methods. Also, the scheme maintains good performance under varying traffic demands.

Xingmin Wang (xingminw@umich.edu), University of Michigan
Yiheng Feng, Purdue University
Yafeng Yin, University of Michigan
Henry Liu, University of Michigan

The max-weight (max-pressure/back-pressure) control for urban traffic networks has drawn much recent attention. It has been proved that under the store-and-forward network model, the max-weight control is a throughput-optimal policy that can stabilize the traffic network when the demand is within the network capacity region. However, most of the existing studies on this topic do not consider the loss of capacity associated with phase switching, which likely undermines the stability of the network. This work proposes a generalized max-weight control including a switching rule that can dynamically adjust the switching frequency according to the congestion level. By introducing a sufficient condition for the network stability, this paper proves that the proposed policy is throughput-optimal under the point-queue model with switching loss. Simulation based on a calibrated network also shows that the proposed control method outperforms the original max-weight control and the actuated control with regard to total delays.