2021

Marija Ostojic, Cubic | Trafficware
Hani Mahmassani (masmah@northwestern.edu), Northwestern University

One of the ways to design more effective signal control strategies is leveraging and synthesizing connected vehicle generated ( CVG)  information to identify traffic states for the controller to operate in a predictive, yet vehicle-actuated manner. The contribution of this paper is twofold: 1) it presents a framework for an advanced, online, signal control logic in a connected environment that utilizes information from CVs to augment high-resolution controller and/or sensor data, and 2) it applies the trajectory analytics to compare the performance of the new controller schemes with CVG data and functionalities relative to conventional, vehicle-actuated, control. The framework puts forward a predictive control logic that schedules phases in an acyclic manner over a variable planning horizon. Phase duration is continually evaluated in response to updated requests for service distributed among equipped vehicles and associated performance indicators. Within the same connected control setup, two measures of effectiveness of a decision were compared to determine the upper bound on the potential effectiveness of a more-responsive control strategy. Finally, the trajectory analytics was used to evaluate the effectiveness of the CV technology-based control scheme against the conventional one. The findings indicate that both control system performance assessment and optimization objectives should change with access to CVG data. Unlike current state of the practice controllers, the developed method is able to handle high and low demand states equally well. The designed connected controller is shown to be robust in handling varying traffic conditions and demand levels.

Zhihong Yao (zhyao@swjtu.edu.cn), Southwest Jiaotong University
Yang Cheng, University of Wisconsin, Madison
Yangsheng Jiang, Southwest Jiaotong University
Bin Ran, University of Wisconsin, Madison

Queue length estimation is of great importance for measuring traffic signal performance and optimizing traffic signal timing plans. With the development of connected vehicle (CV) technology, using mobile CV data instead of fixed detector data to estimate queue length has become an important research topic. This study focuses on real-time queue length estimation for an isolated intersection with CV data. A Kalman filtering method is proposed to real-time estimate the queue length using traffic signal timing and real-time traffic flow parameters (i.e., saturated flow rate, traffic volume, and penetration rate), which are estimated using CV trajectories data. A simulation intersection was built and calibrated using field data to evaluate the performance of the proposed method and the benchmark method. Results show that when the CV penetration rate is at 30%, the average values of mean absolute errors (MAE), mean absolute percentage errors (MAPE), and root mean square errors (RMSE) are only 1.6 vehicles, 20.9%, and 2.5 vehicles, respectively. Besides, the performance of the proposed model is better than the benchmark method when the penetration rate of CVs is higher than 20%, which proves the validity of the proposed method. Furthermore, sensitivity analysis indicates that the proposed method requires a high penetration rate of at least 30%.

Renato Guadamuz, HR Green, Inc.
Houjun Tang, Pennsylvania State University, University Park
Zhengyao Yu, Verizon Wireless
S. Ilgin Guler (sig123@psu.edu), Pennsylvania State University
Vikash Gayah, Pennsylvania State University, University Park

The performance of traffic signal phasing and timing (SPaT) plans are directly related to temporal fluctuations in traffic volumes and their distribution across competing movements at the intersection. A well-timed signal plan generally allocates green times in proportion to the observed volumes. With the advance of data collection techniques and technology, it is possible to obtain and use high-resolution traffic data using detectors at intersections to measure the real-time performance of the SPaT plan to achieve this goal. This study introduces novel green time metrics and methods to evaluate the efficiency of green time allocation along arterials and at individual intersections using automated, real-time, and high-resolution traffic data. An empirical application is presented along the main movement for a radial arterial in Salt Lake City, Utah and a detailed analysis for all the phases and movements for one intersection along that arterial. The proposed methods and results can help identify opportunities to improve the SPaT plan for the entire corridor or movements within secondary approaches by identifying locations and/or time periods where SPaT plans do not align well with observed traffic volumes. These metrics can be automated to provide traffic engineers with an alert as to when a particular intersection should be examined as a candidate for signal retiming, and they can also help identify what changes need to be made with the current SPaT plans.

Zahra Khalilzadeh, Iowa State University
Dorcas Okaidjah, Iowa State University
Christopher Day (cmday@iastate.edu), Iowa State University

Some studies of traffic signal control use simplified phase policies that couple together movements that could otherwise be independently controlled. This study compares more flexible timing, as exemplified by the eight-phase control scheme (but which could be duplicated under stage-based control) against simplified four-phase control schemes that either couple together opposing left and through movements, or each approach (also known as “split phasing”). Three separate experiments are conducted to ascertain the difference between these three types of phase assignments on the signal operation. The results show that in some cases the magnitude of change from implementation of a more flexible phasing scheme can be on the same order of magnitude as an operational innovation, and that the effects of that innovation can be made to appear larger with use of a simplified, less flexible phasing scheme. Results also showed that, for a collection of randomly-generated traffic networks, more flexible phasing schemes consistently yielded better performance within a mesoscopic model. As a counterpoint, the paper concludes by presenting a third simulation experiment that shows an example where the simplified scheme does not yield worse performance. Overall, the results presented here suggest that greater consideration should be made for phase sequence in research on operational innovations at traffic signals.

Chaopeng Tan (tanchaopeng@tongji.edu.cn), Tongji University
Jiarong Yao, Tongji University
Xuegang Ban, University of Washington
Keshuang Tang, Tongji University

Intelligent mobility and connected vehicle technologies have inspired considerable studies on performance evaluation of signalized intersections in recent years. Though, existing studies often require high penetration rates to estimate a single parameter such as queue length, traffic volume and delay, which is not reliable for signal control optimization. In this paper, we proposed a Cumulative Flow Diagram (CFD) estimation and prediction method using sampled vehicle trajectory data for time-dependent fixed-time control intersection. It can not only comprehensively evaluate the performance of fixed-time signalized intersections by estimating the traffic volume, queue length, and delay, but also can be further applied to optimization as the CFD can be updated for candidate signal timing plans. In CFD estimation, a non-parametric method, i.e., Kernel Density Estimation (KDE), is firstly adopted to produce the cumulative arrival curve without any prerequisite assumptions on vehicle arrivals, which enables the method’s applicability to various arrival types. Then, the Huber regression is used to improve the robustness and noise resistance of cumulative departure curve fitting. While in CFD prediction, by updating the cycle arrival time of sampled vehicles, the cumulative arrival curve can be updated correspondingly, hence the CFD for any candidate signal timing plans can be predicted. Evaluation of the proposed method is done using both simulation and empirical data. simulation results indicate that the proposed method performs well in both CFD estimation for the current signal timing plan and CFD prediction for candidate signal timing plans, with an percentage error of no more than 20% even under low penetration rate (5%), showing great potential in signal timing evaluation and optimization. Empirical results show that estimation errors of traffic volume and queue length extracted from the estimated CFD are just 4.2% and 4.0% respectively, as accurate as two state-of-the-art methods, demonstrating feasibility and reliability for real-world application.

Md Asaduzzaman, Louisiana Transportation Research Center (LTRC)
Raju Thapa, Texas Department of Transportation
Julius Codjoe (Julius.Codjoe@la.gov), Louisiana Department of Transportation and Development

Roadway intersections are high risk areas in a roadway network accounting for significant number of traffic fatalities. Of the total fatalities, intersection fatalities account for around 24% each year, with left turning vehicles making up the majority of these crashes. To manage the left turning movements, various left turn signal phasing like permitted, protected only (PO), and protected/permitted left turn (PPLT) are currently in use. Though conversion to PO left turn phase has shown positive effects, their overall effectiveness has not been well established in Louisiana. The paper evaluated the safety effectiveness of PO over PPLT phase in Louisiana. A total of 42 four legged signalized intersections were selected with 21 intersections (treatment) having PO left turn phase at all the approaches and 21 intersection (control) with PPLT phase at all the approaches. Treatment and control sites were selected to closely match with their geometry, traffic volume, and roadway classification. A cross-sectional study design was undertaken to develop a safety performance function, using Negative Binomial model, for total and only left turn crashes at various severity levels. The result shows positive safety benefits of adding PO over PPLT phase, specifically in reducing injury and fatal crashes for both total and left turn crashes (CMF = 0.567 and 0.309 respectively). However, the result for total and left turn crashes was mixed for other severity levels. Nevertheless, installation of PO left turn phase clearly helps to meet the goal of state agencies to reduce zero deaths from roadway crashes.

Peipei Xu, University of Arizona
Xiaofeng Li (xiaofeng.li@hawaii.edu), University of Hawaii
Hyunsoo Noh, Pima Association of Governments
Yao-Jan Wu, University of Arizona

Accurate turning movement counts (TMC) data at signalized intersections is critical to regional transportation planning and simulation modeling. A variety of traffic sensors, configured at intersections for traffic monitoring and signal control, can generate a large amount of real-time high-resolution event-based data from traffic signal controllers but few of these sensors are configured to collect TMC. This paper proposes a methodology to estimate the network-level TMC using event-based data. First, volume-related features are extracted from event-based data, including detector occupancy time, detector-triggered count, green time duration, and left-turn phase type. With these features, a multi-output multilayer neural network model is developed to estimate TMC. In order to further improve estimation accuracy at a network level, infrastructure data and point-of-interest (POI) data are also included as exogenous variables for the proposed model. 84 signalized intersections are chosen from Pima County, Arizona, to calibrate and verify the developed model. The validation results show that the proposed model can accurately estimate TMC, as indicated by the R-squared of 0.93, 0.83, and 0.76 for through movement, left-turn movement, and right-turn movement volume estimation, respectively. This research provides a new possibility of utilizing existing data sources without additional infrastructure and labor costs for transportation agencies to obtain the network-level TMC.

Tianya Zhang (tianya-zhang@utc.edu), University of Tennessee, Chattanooga
Peter Jin, Rutgers University
Thomas Brennan, College of New Jersey
Kelly McVeigh, New Jersey Department of Transportation
Mohammad Jalayer, Rowan University

The traditional project-based signal optimization practices are time-consuming and costly. In contrast to reactive traffic operation and maintenance, data-driven Automated Traffic Signal Performance Measures (ATSPMs) provide a means to proactive management and identify problems on a signalized roadway. The ATSPMs are used as part of an extensive centralized adaptive signal control system that provides a method to assess intersection performance remotely. In a standard system deployment, such as the one used by the Utah Department of Transportation, Linux high-resolution controllers are required to provide the signal event data needed to generate performance metrics and diagrams. Upgrading an adaptive signal infrastructure system with new controllers requires significant investment in funding and labor hours. This paper focuses on customizing an ATSPM system platform so that it can accept signal event log and detector data directly from an adaptive signal control system. This paper provides an efficient and flexible approach to display adaptive traffic signal data without intensive labor configuration and new equipment investment. This approach requires a data processing module that ingests two mainstream data logs from adaptive signal control technology (ASCT) software, InSync and SCATS, into an established ATSPM system in New Jersey. The resulting performance diagrams generated from the ASCT software illustrate the feasibility of the proposed approach. They have led to signal improvement, especially on split failures, pedestrian, and minor street wait time.

Jobaidul Boni (jobaidulalam.boni@mavs.uta.edu), University of Texas, Arlington
Kate Hyun, University of Texas, Arlington
Stephen Mattingly, University of Texas, Arlington

Distracted driving has existed since the invention of the automobile; however, the emergence of the cell phone and especially, smart devices, has created another source of distraction that affects drivers visually, physically, cognitively and audibly. Many studies investigate different factors that influence saturation flow and start up lost time at a signalized intersection, but the impact of new distractions, such as electronic devices and in-vehicle entertainment systems, remains less investigated. This study aims to characterize technology’s influence on driver behavior at intersections and the impact of distractions on startup lost time through a field test conducted at three intersections in Texas. This study uses observational analysis and hypothesis testing to understand distraction behaviors and impacts on individual and aggregated startup lost time. On average, 15% drivers experience distraction during a red indication, and a cell phone distracts more than 60% of these drivers. However, if vehicles locate behind a truck at intersections, 20% drivers are distracted, and 80% of then use a cell phone. More importantly, technology induced distraction, which creates significant intersection delay, is more uncertain and varies from event to event, than non-technology induced delay.  Statistical analysis shows that distraction causes significantly higher headway and total lost time than non-distraction conditions.  This study shows that technology induced distractions from the prevalent use of cell phone or in-vehicle systems makes identification of the variations in different types of distracted driving and their impact on startup lost time and saturation flow critical to properly adapt phase and cycle lengths.

Aobo Wang (aobowang@nevada.unr.edu), University of Nevada, Reno
Zong Tian, University of Nevada, Reno

Signal re-timing has been noted as an important means to improve urban arterial operations. However, most agencies in the U.S. have difficulties in keeping their signal timings up to date due to tight resourcing and staffing. One of the challenges the agencies have faced lies in data collection processes during signal timing that are often costly and laborious. This research aims to provide an alternative method to re-time traffic signals with reduced costs, which is based on leveraging data of controller operations. Abundant operational data logged in the field controllers include information such as historical cycle and phase split times, phase termination logging, and frequency of pedestrian request received, which can be conveniently obtained. Jointly using the controller operational data and a signal timing optimization software, the cycle length, phase splits, offsets, and phase sequences can be properly designed without traffic volume data. Experimental investigations based on hardware-in-the-loop simulation were conducted, which demonstrate that the proposed approach is feasible. In addition, the case study was performed regarding a real-world signal re-timing project in Reno, Nevada. Timing plans designed through the proposed approach and the conventional approach were evaluated, and the results showed that the proposed approach could deliver the similar or even better performance compared to the conventional signal re-timing approaches while a considerable saving in data collection costs could be achieved.

Suhaib Al Shayeb (sma115@pitt.edu), CHA Consulting, Inc.
Nemanja Dobrota, Kittelson & Associates, Inc.
Aleksandar Stevanovic, University of Pittsburgh
Nikola Mitrovic, CHA Consulting, Inc.

Traffic simulation and optimization tools are classified by their practical applicability into two main categories: theoretical and practical. The performance of the optimized signal timing derived by any tool is influenced by how calculations are executed in a particular tool. HCS and Vistro implement the procedures defined in the HCM, thus they are essentially utilized by traffic operations and design engineers. Considering its timing diagram drafting ability, and travel time collection studies, Tru-Traffic is more commonly used by practitioners. All these programs have different built-in objective function(s) to develop optimized signal plan(s) for intersections. In this study, the performance of the optimal signal timing plans developed by HCS, Tru-Traffic, and Vistro are evaluated and compared by using the microsimulation software Vissim. A real-world urban arterial with 20 intersections and heavy traffic in Fort Lauderdale, Florida served as the testbed. To eliminate any bias in the comparisons, all experiments were performed under identical geometric and traffic conditions, coded in each tool. The evaluation of the optimized plans was conducted based on average delay, number of stops, performance index, and travel time. Results indicated that although timings developed in HCS reduced delay, they drastically increased number of stops. Also, Tru-Traffic signal timings, when only offsets are optimized, performed better than timings developed by all of the other tools. Finally, Vistro increased arrivals on green, but it also increased delay. Optimized signal plans were transferred manually from optimization tools to Vissim. Therefore, future research should find methods for electronically transferring optimized-plans to Vissim.

A. M. Tahsin Emtenan (tahsin@umd.edu), CATT Laboratory
Christopher Day, Iowa State University

During oversaturated conditions, common objectives of signal timing are to maximize vehicle throughput and manage queues. A common response to increases in vehicle volumes is to increase the cycle length. Because the clearance intervals are displayed less frequently with longer cycle lengths and fewer cycles, more of the total time is used for green indications, which implies that the signal timing is more efficient. However, previous studies have shown that throughput reaches a peak at a moderate cycle length and extending the cycle length beyond this actually decreases the total throughput. Part of the reason for this is that turning traffic that leaves the through lanes creates gaps in traffic that reduce the saturation flow rate within each lane. There is a relationship between the proportions of turning traffic, the storage length of turning lanes, and the total throughput that can be achieved on an approach for a given cycle length and green time. This study seeks to explore this relationship to yield better signal timing strategies for oversaturated operations. A microsimulation model of an oversaturated left-turn movement with varying storage lengths and turning proportions is used to determine these relationships and establish a mathematical model of throughput as a function of the duration of green, storage length, and turning proportion. The model outcomes are compared against real-world data.

Enrique Saldivar-Carranza (esaldiva@purdue.edu), Purdue University
Howell Li, LSM Analytics LLC
Jijo Mathew, Purdue University
Margaret Hunter, Purdue University
James Sturdevant, Indiana Department of Transportation
Darcy Bullock, Purdue University

ABSTRACT Operations-oriented traffic signal performance measures are important for identifying retiming needs to improve traffic signal operations. Currently, most traffic signal performance measures are obtained from high-resolution traffic signal controller event data, which provides information on an intersection-by-intersection basis and requires significant initial capital investment. This paper proposes using high-fidelity vehicle probe data to produce split-failure, downstream blockage, and quality of progression, as well as traditional Highway Capacity Manual (HCM) Level of Service (LOS). Geo-fences are created at specific signalized intersections to filter vehicle’s waypoints that lie within the generated boundaries. These waypoints are then converted into trajectories that are relative to the intersection. A case study is presented that summarizes the performance of an 8-intersection corridor with 4 different timing plans using over 160,000 trajectories and 1.4 million GPS samples collected during weekdays in July 2019 between 5:00 AM and 10:00 PM. The paper concludes by discussing cloud-based implementation that can cost effectively process over 10 billion records collected statewide each month. Keywords: Traffic Signal Performance Measure, Vehicle Probe Data, Vehicle Trajectory Data, Big Data, Cloud Processing

Sri Rama Bhaskara Kumari Duvvuri (dsrbkumari@hotmail.com), Arcadis US Inc
Montasir Abbas, Virginia Polytechnic Institute

The implementation of a protected/permitted operation for left-turn movement increases safety at an intersection along with improving operational efficiency and reducing traffic delay. There are mainly five different types of PPLT operations used in different states of the United States. They include Circular Green (CG), Flashing Circular Red (FCR), Flashing Circular Yellow (FCY), Flashing Red Arrow (FRA) and Flashing Yellow Arrow (FYA).  It is important to study drivers’ level of perception for each left turn operation to understand advantages and disadvantages of implementation of Protected-Permitted Left Turn (PPLT) signal phasing. The main objective of this paper is to use a real-time driving simulator to study driver behavior at different left-turn operational scenarios. Design of experiments with different control variables such as traffic volume, time of day, speed limit etc., were determined and the proposed scenarios were implemented in the driving simulator. Results were collected and driver safety was analyzed under all scenarios with respect to the parameter "Time to Collision" (TTC). Statistical analysis was performed using the JMP Statistical tool to determine the change in TTC in each scenario. Response surfaces and Prediction Profilers from JMP were analyzed to determine the best suited PPLT operation under different control variables. Graphs were plotted from the results to propose guidelines for different situations. From the results, it was observed that, during the day-time, Flashing Circular Yellow and Flashing Yellow Arrow are most suitable displays. During the night-time, Flashing Circular Red and Flashing Red Arrow are most suitable.

Liang Xia (lxia@my.swjtu.edu.cn), Southwest Jiaotong University
Xiaofeng Li, University of Hawaii
Mohammad Shaon, University of Connecticut
Yao-Jan Wu, University of Arizona
Xinguo Jiang, Southwest Jiaotong University

Traditionally, signal offset optimization for coordinated traffic signals is based on posted speed limits or average speeds among intersections, without consideration of the variations of the travel speed. To achieve more efficient signal coordination along the corridor, an optimization model is developed to optimize offset using real-time speed data collected for the crowdsourced data. The objective of the proposed model is to minimize the average traffic delay of both directions on the corridor. The optimization problem is formulated as an integer programming and the Genetic Algorithm (GA) is utilized to find the best solution. In the numerical exercise, several experiments are first conducted using INRIX speed data to validate the effectiveness and performance of the proposed model. Then, the performance of the proposed model is evaluated for time-of-day (TOD) signal plans as well as the optimal historical data sizes. The results show that: 1) using speed limit to calculate the offset generates the largest traffic delay, while the traffic delay with the use of real-time speed to calculate the offset can save 9.18% and 65.5% compared to those of using average speed and speed limit, respectively; 2) the delay is decreased with the increase in the number of TOD plans and the time consumption to calculate the signal offset is similar regardless the number of TOD plans; 3) five-day of speed data is the optimal size for optimizing the offset and has less than 3% delay difference when comparing with the delay using the offset optimized by the validation dataset.

Qichao Wang (Qichao.Wang@nrel.gov), National Renewable Energy Laboratory (NREL)
Joseph Severino, National Renewable Energy Laboratory (NREL)
Juliette Ugirumurera, National Renewable Energy Laboratory (NREL)
Wesly Jones, National Renewable Energy Laboratory (NREL)
Jibonananda Sanyal, Oak Ridge National Laboratory

Conventional signal timing for arterial usually takes bottom-up approaches. Engineers optimize each individual intersection first and then coordinate them by adjusting the offsets. This work is based on the Virtual Phase-Link (VPL) model, a street traffic model designed for online traffic model predictive control, to obtain a top-down offline arterial signal timing. We Studied the Shallowford Rd. in Chattanooga, TN and found that the inconsistency in intersection capacities along the arterial could lead to some intersections becoming bottlenecks. Signal timing is a significant factor that affect the intersection capacities. We realized that the VPL-based model can guarantee the consistency in intersections along an arterial. We therefore adopted the VPL-based model and developed an offline signal timing optimization approach. The proposed timing derived from the VPL-based offline signal timing optimization showed very good results in simulation. The Chattanooga Department of Transportation adopted the optimized timing obtained from the proposed approach and gave positive feedbacks to the research team. We also collected field experiment data, which demonstrated overall energy reductions and speed improvements on some sections of the Shallowford Rd. arterial. We will continue the experiment when the COVID-19 pandemic impact subsides to have a more robust quantitative evaluation.