Dashboards for Monitoring Congestion and Crashes in Interstate Work Zones

Work zones present challenges to safety and mobility that require agencies to balance limited resources with vital traffic management activities. Extensive literature exists regarding the impact of congestion and recommendations for work zone design to provide safe and efficient traffic operations. However, it is often infeasible or unsafe to inspect every work zone within an agency’s jurisdiction, so it is important to obtain operational feedback regarding congestion and crashes in work zones to prioritize inspection activities. This paper outlines the use of connected vehicle speed data and crash report data to identify operational performance problems in work zones. This is a way to provide feedback to queuing models used to design maintenance-of-traffic (MOT) plans. A weekly work zone report and dashboards were developed for use by the Indiana Department of Transportation (INDOT) for the purpose of assessing and improving both mobility and safety in work zones. The study has developed a mile-hours of congestion graph, frequency of speed delta heat map, congestion profile graph, and the Route Builder interactive application to comprehensively visualize work zone performance. This weekly report provides a mechanism for agency staff to maintain situational awareness of which work zones were most challenging for queues and during what periods those were likely to occur. In one case study, the reports were used to identify and mitigate operational performance problems in a work zone within 4 weeks, reducing congestion and crash rates. The integration of these data provided project managers with quantitative information about traffic mobility and performance of work zones for informed decision-making during the construction season.


Introduction
Maintenance of traffic (MOT) plans are an integral part of traffic operations for any roadway construction, maintenance, or rehabilitation project.The design of an MOT plan often occurs late in the design phase of a project (Figure 1).The process typically involves modeling to predict queue lengths and other impacts on mobility and safety.Most agencies with substantial interstate volumes have strict policies regarding the restriction of traffic on interstates to minimize queuing.Lane closures are often only allowed during certain hours or days with lower traffic volumes to minimize the formation of queues.However, estimating the lane capacity to calibrate a simulation model is a challenge for all states [1].
MOT plans implemented during construction may involve multiple stages depending on the schedule and scope of work activities.Ideally, traffic management personnel would monitor traffic and use the observed impacts to calibrate the queue models and/or make dynamic changes to the MOT plan as needed.
However, with dozens of construction projects underway at any given time, monitoring work zones via regular in-person visits can consume significant manpower.Furthermore, work zones may have subtle changes on a near daily basis that can significantly impact work zone queueing.Indiana Department of Transportation (INDOT) policy states that the maximum allowable queue length is 1.5 miles [2].
The motivations for this work are as follows: • Active monitoring of all active work zones within an agency's jurisdiction may not be feasible due to resource constraints; • Work zone policies are often difficult to enforce due to lack of data; • Models are often used to design MOT plans but are often difficult to validate without real data.
To assist INDOT in dynamic monitoring and assessment of interstate work zones, a weekly work zone report and web-based tools were developed, referred to as dashboards.A dashboard is a visual tool that allows the user to see the status of a system (or a part of a system) in a simple format, similar to how the dashboard of a car allows the driver to easily determine their speed and fuel level.Data from the dashboards give agencies leverage and actionable information to work with contractors on MOT adjustments and improvements.The Federal Highway Administration (FHWA) published the Final Rule on Work Zone Safety and Mobility [1] in September of 2004.In short, this rule states that any roadway project receiving federal funds must have a maintenance of traffic plan.In response, many state and local agencies developed guidelines, policies, or programs for oversight of traffic management plans.The New York State Department of Transportation outlined clear contractual requirements, accident reporting, quality assurance/quality control procedures, etc. in its construction safety and health program [3].The Virginia Department of Transportation developed its own Transportation Management Plan Requirements [4] based on recommendations published in 2005 [5].These requirements apply to all projects within state right-of-way, regardless of funding source.In Washington, DC, a Citywide Transportation Management Plan was deployed to coordinate and analyze work zones and special events [6].INDOT frequently updates its Interstate Highways Congestion Policy, which defines acceptable impacts on traffic, lane closure policy, etc. [2].The purpose of these policies is to maintain capacity and reduce congestion due to work zones.
Evaluation and enforcement are critical in ensuring these policies are upheld and updated.Rouphail, Yang, and Fazio found significant "discrepancies between standards and practice" in their study of short-and long-term work zones [7].In the work zones with such discrepancies, there were higher speed variations between vehicles.Gambatese and Johnson found that traffic management plans were of higher quality and had improved implementation when construction personnel were involved in the design phase and constructability was prioritized [8].To manage compliance and quality, some agencies have developed quality assurance programs and inspection procedures [9].Development and approval of maintenance of traffic plans often involve simulation [10] to assess mobility and safety impacts so effective work zone MOT plans can be designed.The Ohio Department of Transportation uses measured flow data to evaluate and calibrate their queue simulation programs [11].
Performance measures are an integral part of the monitoring and assessment of the impacts of MOT plans.Queue length, travel time, and delay are common performance measures [12] [13].Bourne et al. [14] summarize some of the best practices in work zone assessment, data collection, and performance evaluation.
The State of Virginia has its own performance assessment process [15].Another study considered the effects of quantitative performance measures on the revision of the work zone decision-making process [16].A common theme in all of these initiatives is that actively monitoring work zones and conducting af-American Journal of Operations Research ter-action assessment is important for continued improvement of traffic management and maintenance of traffic plans in the future.
While post-project assessment is important for future decision-making, active monitoring and dynamic management during the course of a work zone can reveal opportunities for improvement in mobility and safety.Real-time measurement of travel time delay can assist motorists in their decision to divert and avoid congestion and could be utilized for contracts with innovative travel time reliability clauses [17].For example, the citywide work zone management and monitoring system developed for Washington DC, included a suite of web-based tools [5].Work zone monitoring tools are valuable to agencies for the dynamic management of traffic at work zones.

Data Sources
This work utilized three data sources: work zone data, connected vehicle speed data, and crash data obtained from statewide crash report database.The following sections detail the nature and use of each data source in this research.Due to the nature of the US transportation system, English units are used as the primary units.

Work Zone Data
For the research presented in this paper, 18 work zones across Indiana were selected (Figure 2) and are a subset of the dynamic list of work zones monitored by these tools.The selected work zones ranged from 1 to 24 miles in length and The number corresponds to the work zone's arbitrary order within that district.
When discussing a particular direction of travel within the work zone, a second letter is added corresponding to the direction of travel.For example, a work zone labeled "C3S" corresponds to the southbound direction of the third work zone in the Crawfordsville district, which is on I-65 between mile posts 197 and 207.

Connected Vehicle Data
Connected vehicle speed data were collected from GPS devices, cellular phones, freight data, or vehicle telematics by a third-party vendor.These data came from 1% -2% of vehicles on interstates in Indiana.Individual vehicle trajectory data were aggregated as minute-by-minute space mean speeds for predefined road segments to preserve driver anonymity.The average road segment length was 0.88 miles.In Indiana, there were approximately 2600 segments covering all of the 2250 directional miles of interstate.Each data point had a timestamp, location, Using these data, performance measures have been created that visually depict the performance of an entire roadway over a period of time.These data have been used for performance measures in Indiana in the Indiana Mobility Report [18] and nationwide in the Urban Mobility Scorecard [19].In Indiana, performance measures for decision-makers were developed using these data [20].These data have also been used for real-time traffic monitoring [21], in which there is a lag of 3 -5 minutes.

Crash Data
Crash data were retrieved from the state crash database and used to supplement S2 American Journal of Operations Research the connected vehicle data.Only crashes that occurred on an interstate in Indiana were used in this study.Personal information about the crash participants and investigating officers, such as names and license plate numbers, were not included in these data.These data did include the following relevant information for each crash: date/time, location, number of injuries/fatalities, primary factor, and manner of collision.

Reports and Dashboards
The work zone reports and associated dashboards have been in use by INDOT since May 2016.The components of the report are compiled into a slide deck and distributed to INDOT traffic and project managers.The components are constructed using a mix of database queries, spreadsheets, and dashboards.Each of the report components will be covered in detail in the following subsections.

Mile-Hours of Congestion
Mile-hours of congestion is a performance metric that combines the number of hours a segment of roadway i belonging to a section S is operating under a critical threshold c v , and the length of the segment i l .The mile-hours of congestion C is calculated using Equation (1).
The time interval j belongs to analysis period T, provided in 1-minute frequency.In this study the critical threshold used is 45 mph.Also included in the work zone report is a summary of the total congestion observed in the INDOT district.In the district-wide view, there are three different mile-hour graphs: the sum of all congestion by speed in the work zones within that district, the sum of all congestion by work zone within that district, and the total congestion on all interstate segments within the district.This view is particularly useful for district managers.The impact of district-or region-wide events, such as weather or holidays, within work zones and the entire district can be observed.Patterns of congestion within the work zone that are also observed within the district can be more easily attributed to non-work zone factors.American Journal of Operations Research

Frequency of Speeddelta
Figure 4, the second page of the report for a work zone, includes two heat maps of speeddelta frequency by day and longitudinal location.The speeddelta is the difference between the average speeds of two adjacent connected vehicle reporting segments.If vehicles are decelerating, the speeddelta will be positive (upstream speed minus downstream speed).In this plot, a threshold of speeddelta greater than or equal to 15 MPH is used so as to eliminate noise from minor changes in speed.Due to the nature of the connected vehicle data segmentation, each horizontal line in the grid represents the point between two adjacent segments.The distance between these points are not to scale in these plots.The darker colored spots represent locations where vehicles slowed down more frequently during that day.

Congestion Profile
The third component of the report for a work zone includes two congestion profiles for the work zone in the current week (Figure 5).The congestion profile was developed as part of the Indiana Mobility Report [18].These plots include  Crashes during the week are called out on the plots by mile marker location and day (color).The severity of the crash is denoted in the callout as property damage only (PDO), personal injury (PI), or fatality (F).
INDOT personnel currently have access to a web-based dashboard that generates these figures automatically based on customizable work zone inputs.

Summary Table
Table 1 shows a summary table, the fourth component, for the current week.In addition to columns containing information about the work zone and location, this tables includes the following columns: • Queueing ≥ 5 mi (hr): number of hours when there was a queue of length greater than or equal to 5 miles within or overlapping the work zone.This is a good measure of the duration of severe traffic incidents.
• Queueing upstream of WZ (hr): number of hours when there was a queue extending upstream of the work zone boundary.This measure is important to traffic and project managers in regard to the placement of advance warning signs and queue length modeling.
• Mile-hours < 45 MPH (week): mile-hours of congestion in the work zone for the whole week.This measure is useful in comparison to previous weeks.
• Mile-hours < 45 MPH (worst day): mile-hours of congestion in the work zone for the "worst" day.This measure is useful for determining the impact of recurring congestion relative to individual incidents.
• PDO Crashes: number of property-damage-only crashes within the work zone.
• PI Crashes: number of personal injury crashes within the work zone.Fatal crashes, due to their rarity, are included in this number but are called out in the table with an "*" and in the notes.
• BOQ Crashes: number of back-of-queue crashes within or upstream of the work zone.A back-of-queue crash is a crash that occurs at a shockwave boundary of a queue that exists within or overlaps with the work zone.

Route Builder
The previous sections discuss aggregate measures and visualizations of the congestion within the work zone.However, it is often useful to view the data in an   Active monitoring of work zones using these tools allow for more efficient use of time and resources.Managers can make informed decisions regarding the deployment of assets, enforcement, or the disbursement of information to the public.For example, INDOT policy [6] states that the maximum allowable queue length is 1.5 miles.Table 2 summarizes data from the work zone reports for 5 months (April-August) in the 2017 construction season.These reports can be and have been used to highlight work zones that are underperforming in regards to mobility and traffic operations.

Conclusions
Connected vehicle data and crash data were utilized to develop a weekly work

Figure 1 .
Figure 1.Work zone maintenance of traffic plan development flow chart.
included a variety of construction activities.INDOT personnel provide the data for and select all work zones based on expected congestion, publicity, and duration.The selected work zones are divided by INDOT district, route, direction, and start/end mile posts.The callouts in Figure 2 define the shorthand label used for each work zone in this paper.The first letter corresponds to the district.

Figure 2 .
Figure 2. Map of selected work zones.

Figure 3 ,
Figure 3, the first page of the report for a work zone, includes two plots of mile-hours by day, one for each direction.A mile-hour is a measure of congestion that is temporally and spatially weighted and based on the average operating speed from the connected vehicle data.Each column in the graph represents the total number of mile-hours of operation within each speed bin in one day within the work zone.It does not include the congestion that extends or occurs outside of the work zone.These plots allow personnel to view overall performance and quickly identify days or weeks that had more severe congestion.The 4 weeks prior to the current week are included to provide context and to show any emerging trends.

Figure 3 .
Figure 3. Work zone mile-hours of congestion plots.

Table 1 .
Work zone summary table.