Howell Li

Performance Characterization of Arterial Traffic Flow with Probe Vehicle Data

Extensive literature in the adaptive control field uses local detection available from the traffic controller as input to various control models to adjust splits, cycle lengths, and offsets. All these models have implicit control objectives, which include facilitated progression, minimized stops, minimized delay, and equitable allocation of green time. Enormous opportunities exist to incorporate probe data into the decision process with respect to when and where adaptive control can be used and which operating objectives are most applicable to a corridor as well as to an outcome assessment tool to evaluate the effectiveness of adaptive control. The research reported in this paper compared how probe data sources could be used to identify appropriate adaptive control objectives and to assess the performance of adaptive systems. Four case studies demonstrated how travel time data could be used to evaluate existing conditions, to evaluate the outcome of a traditional signal retiming, and to assess the feasibility of adaptive control opportunities. Currently, the richest probe data sets are provided by agency-installed equipment. Given the increasing penetration of crowd-sourced probe data devices and the onset of connected vehicle infrastructure, however, these sources could provide similarly rich data. This paper recommends that commercial data providers begin to develop more detailed base maps. These maps would provide richer probe data information, such as hour-by-hour statistical distributions and approach delay for signalized arterials for which the segments did not span multiple intersections. This recommendation should motivate agencies to develop more detailed specifications for probe data that will better serve their needs.

Graphical Performance Measures for Practitioners to Triage Split Failure Trouble Calls

Detector occupancy is commonly used to measure traffic signal performance. Despite improvements in controller computational power, there have been few innovations in occupancy-based performance measures and little integration with other data. This paper introduces and demonstrates the use of graphical performance measures based on detector occupancy ratios to verify potential split failures and other signal timing shortcomings reported to practitioners by the public. The proposed performance measures combine detector occupancy during the green phase, detector occupancy during the first 5 s of the red phase, and phase termination cause (gap out or force-off). They are summarized by time of day to indicate whether the phase is undersaturated, nearly saturated, or oversaturated. The graphical performance measures and related quantitative summaries provide a first-level screening and triaging tool to help practitioners assess user concerns about whether sufficient green times are being provided to avoid split failures. They can also provide outcome-based feedback to staff after split adjustments have been made to determine whether operation improved or worsened. The paper demonstrates how the information was used to make an operational decision to reallocate green time that reduced the number of oversaturated splits on minor movements from 304 to 222 during a Thursday 0900 to 1500 timing plan and from 240 to 180 during a Friday 0900 to 1500 timing plan.

Performance Ranking of Arterial Corridors Using Travel Time and Travel Time Reliability Metrics

Performance measures are essential for managing transportation systems and demonstrating agency accountability. Probe vehicles are an effective means for gathering vast amounts of information about highway networks. This paper presents a scalable methodology for analyzing arterial travel times that considers both the central tendency and the reliability of the travel time. A pilot analysis was carried out for 28 arterials with a total of 341 signalized intersections across Indiana. Starting from individual minute-by-minute speed records, the data were converted into travel times and aggregated into time series cohorts that correspond to typical traffic signal time-of-day periods. The data were normalized for the ideal travel time (based on the speed limits on each route) to account for individual route lengths and speeds. The data were compiled for all Wednesdays from January through July 2014 for investigation of arterial characteristics. The results show that a greater density of traffic signals on a route loosely corresponds to higher average travel times and less reliability. A composite index incorporating both the average values and reliability characteristics of travel time is developed and is used to rank the arterials by performance.

Performance Measures for Optimizing Diverging Interchanges and Outcome Assessment with Drone Video

Diverging diamond interchanges (DDIs) are an emerging interchange configuration that eliminates the need for left-turn phases in conventional diamonds and may be less expensive to construct than some alternative geometries. This paper examines signal timing for DDIs. DDI signal timing typically has used a two-phase configuration that reflects the two competing movements at the crossover points at each inter section of the DDI. This configuration inherently contains some inefficiency: (a) there is potential for internal queuing under two-phase configuration and (b) it is possible for the inflow demand to exceed outflow capacity of the interchange. This paper uses high-resolution event data to develop performance measures for evaluating operations at a DDI in Salt Lake City, Utah. Alternatives to the existing signal timing within the two-phase configuration are modeled and tested with a field deployment. The field deployment demonstrated the ability to prioritize ramp or through vehicles within the two-phase configuration. Additionally, a new three-phase configuration was developed and deployed to address the internal queuing that occurs with two-phase timing. With this new configuration, the flows from one DDI intersection to the other are balanced, and progression within the DDI is improved. With the implementation of the three-phase configuration, the percentage of vehicles arriving on green at the heaviest internal movement within the DDI increased from 53% to 92%. To illustrate these performance measures and improved DDI operation qualitatively, a video from a tethered unmanned aerial vehicle demonstrated the vehicle arrival characteristics by overlaying vehicle detection and signal state graphics on the video.

Shock Wave Boundary Identification Using Cloud-Based Probe Data

An important component of active management of freeways is the systematic identification of recurring and nonrecurring congestion, particularly the location of the shock wave boundary between the two flow regimes. In the past five decades, many publications have described point-based detection models. The emerging widespread availability of true space mean speed data obtained from probe vehicles greatly simplifies the incident detection problem. This paper describes the use of cloud-based crowdsourced probe data for detecting the boundary between uncongested and congested conditions. Time-stamped freeway segment speed data can be retrieved from a cloud source with standard web communication and data protocols on a predetermined time interval. After data retrieval, differences in speeds of adjacent roadway segments are computed for all segments across the state. The calculated differences are called the “delta speeds.” A threshold is then set for the computed delta speeds, and any record that surpasses the threshold triggers a warning via an incident management website to alert traffic management personnel. The delta speed data can also be aggregated over time for strategizing the management of shifting work zones or making capital investment decisions. This study used delta speed values to examine an incident on I-69 in northeast Indianapolis, Indiana. Data within a 5-h period were collected to analyze the dynamics of the back of the queue during reduced capacity caused by incident investigation. In addition, the area of analysis was expanded to 150 mi of I-69 over a 1-month period for finding locations with recurring backs of queues.

High-Resolution Controller Data Performance Measures for Optimizing Diverging Diamond Interchanges and Outcome Assessment with Drone Video

Diverging Diamond Interchanges (DDIs) are an emerging interchange configuration that eliminates the need for left-turn phases in conventional diamonds and may be less expensive to construct than some alternative geometries. This paper examines signal timing for DDIs. To date, DDI signal timing has typically used two-phase configuration reflecting the two competing movements at the cross-over points at each intersection of the DDI. This configuration inherently contains some inefficiency: i) there is potential for internal queuing under two-phase configuration, and ii) it is possible for the inflow demand to exceed outflow capacity of the interchange. This paper uses high-resolution event data to develop performance measures to evaluate operations at a DDI in Salt Lake City, Utah. Alternatives to the existing signal timing within the two-phase configuration are modelled and tested with a field deployment. The field deployment demonstrated the ability to prioritize ramp or thru vehicles within the two-phase configuration. Additionally, a new three-phase configuration was developed and deployed to address the internal queuing that occurs with two-phase timing. Under this new configuration, the flows from one DDI intersection to the other are balanced and progression within the DDI is improved. By implementing the three-phase configuration, the percent of vehicles arriving on green at the heaviest internal movement within the DDI increased from 53% to 92%. To qualitatively illustrate these performance measures and improved DDI operation, a video from a tethered unmanned aerial vehicle (UAV) was prepared that demonstrates the vehicle arrival characteristics by overlaying vehicle detection and signal state graphics on the video. Hainen, Stevens, Day, Li, Mackey, Luker, Taylor, Sturdevant, Bullock 3 INTRODUCTION The Diverging Diamond Interchange (DDI) is an innovative interchange geometry whose use is increasing. Typically, DDI signal timing has consisted of a relatively simple two-phase configuration at each intersection. Most of the published literature on DDIs has focused on geometrics, capacity, and safety, with very little published on signal timings. This paper summarizes the DDI literature and describes the application of high-resolution controller data to develop performance measures to optimize DDI timing. BACKGROUND The diverging diamond interchange was developed by Chlewicki from 2000-2003 and was first deployed in the U.S. at I-44 and Kansas Expressway (SR 13) in Springfield, Missouri (1) (2). Bared et al. reported increased capacity at DDIs compared to regular diamond interchanges under various simulations mainly due to the elimination of the left-turn phase which reduces signal timing to a two-phase configuration (3). Several other papers used microsimulation-based evaluations (4) and all reported that for a majority of the cases that DDIs performed better than conventional diamond interchanges, single-point urban interchanges (5), and other alternative intersections often due to the two-phase signal operations (6). In recent work, the HCM delay formulas have been adapted to DDIs (7), simulation packages have been created (8), selection tools have been developed (9), split optimization software made available (10), and critical lane volume analysis tools implemented (11). Cycle length has been investigated (12) and timing the DDI with overlaps instead of a traditional two-phase configuration has been considered (13). In summary, existing work has focused on simulation-based comparisons of the DDI to other operational types. This study reports on the application of high resolution traffic signal controller performance measures to assess DDI field operations and identify opportunities to improve signal timing. The evaluation is carried out during peak and off-peak periods for a DDI in Salt Lake City, Utah. PERFORMANCE MEASURE APPROACH One of the challenges of DDI timing is that under some demand scenarios there is an imbalance between upstream and downstream capacity. Given the relatively limited number of field deployments, there are opportunities to identify and establish performance measures that can be used for optimizing DDI timing efficiencies. This paper documents the following contributions  First, performance measures are developed for the DDI. Graphical tools are applied to the case of the DDI to evaluate how the interchange is performing.  The performance measure data is further applied to explore the optimization of traffic flows between the two intersections. By associating traffic signal states with vehicle detection times, it is possible to identify traffic origins (ramp or mainline) and optimize the total number of arrivals on green.  To address the imbalance between upstream inflow demand and downstream intersection outflow capacity that is inherent in the two-phase configuration, a three-phase configuration which features a “hold-back” phase is proposed and tested in the field. The new configuration improves traffic flows within the DDI by balancing green times at each intersection so that periods of inflow are aligned with compatible greens, and eliminating stops and queuing in the middle of the interchange. Hainen, Stevens, Day, Li, Mackey, Luker, Taylor, Sturdevant, Bullock 4  Lastly, video obtained from a tethered unmanned aerial vehicle (UAV) was prepared that demonstrates the vehicle arrival characteristics by overlaying vehicle detection and signal state graphics on the video. SITE CHARACTERISTICS The DDI at SR-201 and Bangerter Highway in Salt Lake City, Utah was the study interchange for this work. This interchange was formerly configured as a traditional diamond, but experienced substantial congestion during peak periods. Progression along the arterial (14) is a critical consideration at this location due to the high volume of truck traffic to/from the area north of the intersection (adjacent signal approximately 1,000 feet away), as well as the closely spaced signalized intersection to the south of the intersection (adjacent signal approximately 500 feet away). A DDI option was selected as the most cost effective upgrade to mitigate congestion. Figure 1 shows the lane configuration and phasing for the interchange that was re-constructed in 2011. At the north intersection (WB ramp), there are two sources for SB traffic:  External Southbound Thru (Overlap-E). Overlap-E is driven by Ø5 (phase 5) and Ø6.  Westbound Left Movement (Overlap-H). Overlap-H from the ramp is driven by Ø8. At the south intersection (EB ramp), there are two sources for NB traffic:  External Northbound Thru (Overlap-A). Overlap-A is driven by Ø1 and Ø2.  Eastbound Left Movement (Overlap-D). Overlap-D from the ramp is driven by Ø4. Each ramp intersection includes two movements that provide separate inflows to the other intersection. In Figure 1, the black arrows (OL-E and OL-A) are the movements from the external thru movements along Bangerter Highway, and the gray arrows (OL-H and OL-D) are the movements carrying traffic off the ramps. All ramp movements are signalized and turning left on red is not permitted. A 60-second cycle length was used with a 5-second offset between the rings. Ring 1 handles the south intersection and Ring 2 handles the north intersection. The even-numbered phases (2,4,6,8) each have 23-second splits. The odd-numbered phases (1,3,5,7) each have 7second splits. The sole purpose of the odd-numbered phases is to provide additional clearance time to allow the vehicles in the cross-over area to clear the off-ramp conflict points before allowing ramp traffic to proceed. Backup prevention rules were programmed in the controller to ensure that these clearance phases are always served. In some controllers, the same behavior could possibly be achieved without using separate phases by using leading overlaps with red time. Although the ramp and thru movements are controlled by overlaps at this location, the existing timing is effectively the same as “two-phase” operation, since each intersection effectively alternates between two states (northbound or southbound at each intersection). The north intersection alternates between Overlaps G and E while the south intersection alternates between Overlaps A and C. Changing the number of “phases” is considered later in this paper. The locations of the advance detectors are shown in Figure 1. Radar detection was used. The posted speed limit on the bridge is 40 MPH. Hainen, Stevens, Day, Li, Mackey, Luker, Taylor, Sturdevant, Bullock 5 PERFORMANCE MEASURES In previous studies, Indiana has leveraged high-resolution event-based data (15) to develop performance measures for engineers to better understand arrival characteristics and other operational characteristics at intersections (16). The two common performance measures used are the Purdue Coordination Diagram (PCD) and the flow profile diagram. Purdue Coordination Diagram The PCD (17) is a graphical tool for visualizing progression between intersections. Each vehicle arrival time at the stop bar is plotted by the time of day along the horizontal axis and by time in cycle along the vertical axis. Moving from the x-axis upward, the beginning of the cycle is the beginning of red; the next event is the start of green (green line), and the final event is the end of green (red line). Vehicle arrivals represented by dots below the green line represent arrivals on red, while dots above the green line represent arrivals on green. The greater the proportion of dots above the green line, the higher the percent of arrivals on green and the better the progression of vehicles. Figure 2a and Figure 2b are further enhanced by coding the arrivals according to their originating phase. Figure 2a is the PCD for the northbound movement at the north intersection. The black dots (northbound thru vehicles) are located in a dense band immediately following the start of green. This suggests that at the start of green, t

Characterizing Signalized Intersection Performance Using Maximum Vehicle Delay

Average delay is perhaps the most commonly used measure for characterizing the performance of signalized intersections. Current methodologies for estimating the average delay rely on the use of models based on volumes and green times. In practice, it is challenging to develop such real-time measurements of delay, due to the difficulty of accurately measuring vehicle arrivals and departures. However, measuring wait time after the first vehicle arrival during the red interval can be an important performance measure for low and moderate volume conditions. The maximum wait time performance measure provides an upper bound, or maximum, on individual vehicle delay during a given cycle and facilitates comparison between different types of operation.

Integrating Traffic Signal Performance Measures into Agency Business Processes

INTEGRATING TRAFFIC SIGNAL PERFORMANCE MEASURES INTO AGENCY BUSINESS PROCESSES This report discusses uses of and requirements for performance measures in traffic signal systems facilitated by high-resolution controller event data. Uses of external travel time measurements are also discussed. The discussion is led by a high-level synthesis of the systems engineering concepts for traffic signal control, considering technical and nontechnical aspects of the problem. This is followed by a presentation of the requirements for implementing data collection and processing of the data into signal performance measures. The remaining portion of the report uses an example-oriented approach to show a variety of uses of performance measures for communication and detector system health, quality of local control (including capacity allocation, safety, pedestrian performance, preemption, and advanced control analysis), and quality of progression (including evaluation and optimization).

Revisiting the Cycle Length—Lost Time Question With Critical Lane Analysis

During oversaturation, a popular objective in traffic signal operations is to maximize throughput to keep traffic moving. As cycle lengths are increased, the proportion of lost time used to transition between signal phases is reduced. This factor is often a rationale for programming long cycle lengths into signal timing plans. An investigation of the impact of cycle length revisited the concept of the use of critical lane analysis to calculate throughput and applied the technique to data collected at an oversaturated intersection in Indianapolis, Indiana. Traffic volumes were measured for 10 weeks while various cycle lengths, ranging from 80 to 135 s, were tested at the intersection. During saturated conditions, no clear increase in the sum of critical lane throughput was observed, even when the cycle length increased by more than 50%. At 135 s, there was a slight reduction in the total critical lane sum volume. These findings concur with a recent study by Denney et al. The decrease in throughput during the longer cycle lengths is attributed to the reduction of saturation flow during long green times. Possible results of use of a time-dependent saturation flow rate are discussed. Additionally, critical lane analysis may have applications to evaluation and ranking of intersections within corridors as under, near, or over saturation.

Longitudinal Performance Measures for Assessing Agencywide Signal Management Objectives

A fundamental concept of system engineering is defining operating objectives. Then the appropriate strategy and the performance measure for assessing how well the system is working are selected. Two operating objectives are broadly shared by many agencies: reliable communication with the traffic signal infrastructure and good allocation of green times. The first objective is oriented toward asset management and the second toward operational efficiency. However, substantial synergies exist between the objectives. The proposed performance measures can be used for evaluating whether these objectives are being met. The quality of communication is evaluated by an examination of connectivity and data completeness and is based on an open standard telecommunications model. Opportunities to improve green time allocation are detected through an identification of operating patterns in which some phases routinely force off and other phases gap out in the same cycle. The performance measures are presented in an easy-to-understand visual format that practitioners can use. One outcome of this system engineering approach is the realization that many objectives can be achieved with a high-latency, low-bandwidth communication infrastructure if the appropriate processing techniques and performance measures are clearly articulated. The architecture and performance measures are explained in the context of a deployment of 122 intersections geographically distributed throughout the state of Indiana; data were collected during a 2.5-month period with a communication infrastructure that included commercial cellular data and agency-owned fiber.