EVALUATION OF AUTOMATED WORK ZONE INFORMATION SYSTEMS

TABLE OF CONTENTS CHAPTER Page

1. INTRODUCTION .................................................. 1
Present Situation .................................................. 2
Structure of the Report .............................................. 3

2. BACKGROUND .................................................... 5
Traffic Flow Terms................................................. 5
Work Zone Congestion Mechanisms...................................... 9
Data Collection Methods.............................................. 9

3. RESEARCH PROCEDURES........................................... 11
Comparing AWIS with Observations .................................... 11
Site Selection for Capacity Study ....................................... 11
Equipment Selection ............................................... 12
Field Data Collection............................................... 13
Data Reduction .................................................. 19

4. DATA ANALYSIS AND RESULTS ...................................... 23
Comparison of AWIS Display with Observed Backups ......................... 23
Formatting Data to Identify Volumes Associated with Backups.................... 23
Analysis of Data to Identify Volumes Associated with Backups ................... 25
Flow Rates Associated with Backups .................................... 30

5. CONCLUSIONS ................................................... 33
Previous Research ................................................ 33
Research Procedures ............................................... 33
Observations .................................................... 34
Conclusions..................................................... 35

REFERENCES....................................................... 37

APPENDIX 1 Typical Speed Data Reduced from Lidar Gun ...................... 39
APPENDIX 2 Typical Volume Data Reduced from Video ........................ 41

LIST OF FIGURES

Number Title Page 3-1 Typical Station Setup............................................... 14 3-2 Study Site with Respect to AWIS Sensors at Kerr Road ........................ 15 3-3 Stations at Galloway and County Line.................................... 16 3-4 Typical Setup on Scaffolding.......................................... 16 3-5 Relative Conspicuity of Data Collection Setup to Approaching Drivers . . . . . . . . . . . . . . 17 3-6 Station 3 Partially Hidden Behind a Portable Sign ............................ 18 3-7 Station 4 at Shearerville Concealed by the Parked Vehicle....................... 18 4-1 AWIS Agreement with Field Observations................................. 24 4-2 Example Trellis Plot ............................................... 25 4-3 Volume by Speed at Station 1 Conditioned on Backups at the Other Three Stations . . . . . . 26 4-4 Volume by Speed at Station .......................................... 26 4-5 Robust Regression Fit for Volume by Speed at Station 1........................ 27 4-6 Moving-Sum-of-Volume ............................................ 28 4-7 Modified Graphs for Shearerville Site .................................... 29 4-8 Probability of Backup Occurred with 1-min Moving Volume Interval . . . . . . . . . . . . . . . 31 4-9 Probability of Backup Occurred with 3-min Moving Volume Interval . . . . . . . . . . . . . . . 32 4-10 Probability of Backup Occurred with 5-min Moving Volume Interval . . . . . . . . . . . . . . . 32

LIST OF TABLES

Number Title Page
2-1 Summary of Observed Capacities for Long Term Work Zones..................... 8
2-2 Summary of Observed Capacities for Work Zone Types ......................... 8
2-3 Capacity Values for Long Term Construction Work Zones ....................... 8
3-1 Distances from the Taper Begin to the Stations .............................. 14
4-1 AWIS Agreement with Field Observations................................. 24

EVALUATION OF AUTOMATED WORK ZONE INFORMATION SYSTEMS

by
Melissa S. Tooley, Ph.D., P.E., J. L. Gattis, Ph.D., P.E., Rajeshkumar Janarthanan,
Mack-Blackwell National Rural Transportation Study Center, and
Lynette K. Duncan, Center for Statistical Consulting,
University of Arkansas

CHAPTER 1
INTRODUCTION

In the late 1990s, Arkansas embarked upon an ambitious project to rehabilitate and reconstruct a large portion of its existing Interstate highway network. Having a significant portion of the freeway system under construction within the span of a few years resulted in a heightened probability that a motorists would encounter lane closures and the congestion that can occur when one of the two through lanes in a direction is closed to traffic.

This research project, sponsored by the Arkansas State Highway and Transportation Department (AHTD) and the Mack-Blackwell Rural Transportation Center (MBTC), was conducted at long-term rural Interstate highway work zones where two lanes of traffic in a given direction were reduced to one lane. Originally, the study had two objectives: 1) to evaluate the effectiveness of AWIS (Automated Work Zone Information System) technology in managing congestion, improving safety, and providing traveler information, and 2) to evaluate the public’ s perception of the technologies in use. After consultation with AHTD, it was decided during the course of the project to focus on the first objective. It became apparent that the data collection would be more challenging than anticipated due to several factors, including cancellations of data collection trips due to such things as equipment failure, weather, and unpredictable construction schedules. After preliminary results were obtained from field data collection, additional trips were scheduled to obtain more significant results. These exhausted the resources available for the project, but resulted in high quality data for analysis. In order to evaluate the effectiveness of AWIS technology, the following tasks were accomplished.

1. Compare the traffic backup indications obtained from an Automated Work Zone Information System (AWIS) with the backups observed in the field, and thus assess the effectiveness of AWIS in determining the presence of backups and notifying motorists.

2. Determine at what volumes traffic in Arkansas’ rural freeway work zones began to experience

backups and congestion. The work was conducted by the Mack-Blackwell Rural Transportation Center (MBTC) at the University of Arkansas.

PRESENT SITUATION

Motorists traveling through a work zone are faced with both routine and unexpected driving tasks, coupled with additional distractions provided by the work zone. Distractions may include slow or stop-and-go traffic, narrow or a reduced number of lanes, and entering and exiting construction vehicles. In addition, a portion of the driver’ s attention may be diverted to the roadwork activity being performed outside the travel way.

Some work zones require the closure of one or more through-traffic lanes. When traffic volume exceeds the capacity of the merging point or of the reduced number of lanes, the resulting congestion can lead to the formation of queues, which results in delays. As upstream vehicles traveling near 110 km/h (70 mph) approach a slow-moving or stopped queue on a freeway, the potential for traffic crashes increases.

Various types of traffic control devices have been employed in an attempt to reduce hazardous conditions associated with the work zones. Despite significant applications of work zone traffic control, work zones are known for exhibiting heightened crash rates. As a result, transportation agencies at all levels have been searching for new methods to alert and warn drivers of potential hazards. Recent years have seen a renewed interest in work zone traffic control devices and related research.

An AWIS provides useful real-time information about traffic conditions to motorists as they approach and pass through a work zone. The intended purposes of providing the real-time information in a work zone include informing drivers that they can expect congestion and delays ahead, and improving safety for both motorists and construction personnel. A successful AWIS installation will benefit motorists by reducing driver frustration and increasing safety.

An AWIS has been employed at some freeway work zones in Arkansas. The system for westbound traffic on I-40 east of North Little Rock included seven traffic sensors and a changeable message sign (CMS). The RTMS (Remote Traffic Microwave Sensors) communicate by radio with a central controller and, based upon real-time speed or queue measurements, choose the appropriate preset messages to display on a CMS for approaching motorists. The AWIS system was the CHIPS^® (Computerized Highway Information Processing System), developed by ASTI Transportation Systems (Tudor, Meadors, and Plant 2003) .

This project evaluated the effectiveness of the AWIS technologies currently being used, by comparing the AWIS indications of when backups were present with field observations of backup occurrence. A second question addressed by this project was what traffic flow rates lead to congestion at freeway work zones where two lanes were reduced to one.

STRUCTURE OF THE REPORT

Chapter 1 introduces the project and describes the basic concept of the AWIS application and work zone congestion. It also summarizes the structure of the report.

Chapter 2 of this report presents the basics of freeway traffic flow and some definitions related to freeway congestion. It also gives information about the equipment used to collect certain types of work zone research data.

Chapter 3 begins with an overview of the equipment used and the site selection process. The methodology implemented to set up instruments in the field and techniques carried out to collect the traffic data are discussed. Data reduction is also described in this chapter.

The analyses performed on the field data is documented in Chapter 4. The various methods followed to plot the graphs relating the speed, volume, and time are discussed here.

Chapter 5 presents a summary of the findings and conclusions from this research. It also includes a discussion of practices which could be modified to reduce the frequency of backups.

blank page

CHAPTER 2
BACKGROUND

A literature review was conducted to find information related to work zone information systems, data collection methods and procedures, equipment requirements, and other flow-related information for the project. This review also explored the various methods to estimate the traffic-carrying capacity of work zones and to analyze the behavior of traffic in work zones.

TRAFFIC FLOW TERMS

Topics such as flow, density, speed, capacity, and level-of-service help define the quantity and quality of traffic flow. Flow

Flow (often identified as the variable q) is most commonly defined as the quantity or rate of traffic moving along a section at a given time. Flow varies over time. It can indicate either the number of vehicles being processed or the number arriving to be processed per unit of time.

Common units of flow include vehicles per hour (veh/h), vehicle per hour per lane (veh/h/ln), or passenger car equivalents per hour (pce/h). Sometimes the term “ volume” is used to indicate the number of vehicles passing a point in an hour or a day, while “ flow rate” means the number of vehicles in a shorter time, but expressed in terms of vehicles per hour. For instance, 100 vehicles in 15 minutes is equivalent to a flow rate of 400 veh/h.

As shown in Equation 2.1, flow is the inverse of headway, the time between the passage of the front bumpers of consecutive vehicles. Equation 2.1

Density

Vehicle density is the number of vehicles per length of road and is typically noted as the variable

k. The units used in denoting density are vehicles per kilometer (veh/km) or mile (veh/mi), or vehicles per lane-km (veh/ln-km) or lane-mi (veh/ln-mi).

The Highway Capacity Manual (HCM) defines density as “ the number of vehicles on a roadway segment averaged over space”. Mathematically, it is the inverse of the distance from front bumper to front bumper of consecutive vehicles (see Equation 2.2).

Equation 2.2

Speed

Vehicle speed is usually indicated by the variables v or s. Common units of measure are kilometers per hour (km/h) or miles per hour (mph).

Speed can be measured as either space-mean-speed or time-mean-speed. The HCM defines space-mean-speed (SMS) as “ the average speed of the traffic stream computed as the length of the highway segment divided by the average travel time of the vehicles to traverse the segment.” Time-mean-speed (TMS) is “ the arithmetic average of individual vehicle speeds passing a point on a roadway or lane, in miles per hour.” Relating Flow, Density, and Speed

Flow is a function of vehicle density and the speed of the traffic flow. Equation 2.3 shows the relationship used to estimate flow. Equation 2.3

This flow-density-speed relationship employs the space-mean-speed, as opposed to time-mean-speed.

Capacity

Capacity is the maximum number of vehicles a given facility can accommodate and still not have any type of failure (such as excessive delay, stop-and-go congestion, etc). The HCM definition for capacity is “ the maximum sustainable flow rate at which vehicles or persons reasonably can be expected to traverse a point or uniform segment of a lane or roadway during a specified time period under given roadway, geometric, traffic, environmental and control conditions.” Capacity Values

The standards for calculating flow and capacity are given in the Highway Capacity Manual. The HCM tables list the theoretical capacity of a freeway as 2400 passenger cars per hour per lane (pcphpl) if the free flow speed is 120 km/h (75 mph) and 2250 pcphpl if the free flow speed is 90 km/h (55 mph). The actual capacity value is sensitive to the geometric characteristics of each site. Traffic characteristics such as the percentage of trucks, the directional split, the presence of vehicles on the shoulder, or even the occurrence of a potential incident during the dense flow conditions have a great impact on roadway capacity. It is also dependent on the time interval for which data are collected and from which the capacity is calculated (Polus, Craus, and Livneh 1992). Population Effects on Capacity

Driver population refers to the mix of driver types in a traffic stream by trip purpose. Driver population is a significant factor affecting freeway capacity (Al-Kaisy and Hall 2001).

Adjusting Capacity for Effects of Heavy Vehicles

Passenger car equivalents (pce) take into account the increased impact of larger vehicles (trucks, buses, and recreation vehicles) in comparison to a passenger car. The volume of these heavy vehicles is multiplied by an adjustment factor to account for their larger impact while negotiating the road section. Short-Term Work Zone Capacity

The capacity of a work zone can vary from that of a “ regular” freeway section. In case of short term work zones, the capacity is nominally defined as 1600 pcphpl, regardless of the lane closure configurations (Krammes and Lopez 1994). The HCM suggests that for some types of activities, the capacity may be higher or lower by as much as 10% (HCM, 2000). The HCM also includes an equation for calculating the capacity at short term work zones (Equation 2.4).

C_a = (1600 + I – R) * f_hv * N Equation 2.4 where, C = adjusted mainline capacity (veh/h)

a

f_hv = adjustments for heavy vehicles

I = adjustment factor for type, intensity and location of the work activity

R = adjustment for ramps

N = number of lanes open through the work zone Long-Term Work Zone Capacity

The 1997 edition of the HCM presented capacity values for long term work zones as a function of the number of lanes under normal operation versus number of lanes open during construction. Table 2-1 gives average capacities for different lane closure situations. The variation in the capacity value can be significant and hence HCM suggests that analysts use less than average values to reduce the risk of capacity overestimates. To enable the reader to calculate lower values, the HCM provided a cumulative distribution plot of capacities. Note that work zone capacity is affected by the presence of merging, diverging or weaving movements; grades; alignment; trucks; and other factors.

The HCM 1997 edition also gave capacity values for different kinds of freeway work zones. These are presented in Table 2-2.

The capacity for long term construction work zones are given in the Table 2-3. These values were given by Dudek in his research on work zone capacity. The HCM also agrees with this value, which is included in the 2000 edition.

TABLE 2-1 Summary of Observed Capacities for Long Term Work Zones

Number of lanes			Number of	Average Capacity
Normal		Open	Studies	(veh/h) (vphpl)
3	1		7	1170	1170
2	1		8	1340	1340
5	2		8	2740	1370
4	2		4	2960	1480
3	2		9	2980	1490
4	3		4	4560	1520

TABLE 2-2 Summary of Observed Capacities for Work Zone Types

Capacity Values for normal 2 lane
Type of Work freeway when 1 lane is closed

Median barrier/guard rail	1500 veh/h
Pavement repair	1400 veh/h
Resurfacing, asphalt removing	1200 veh/h
Striping, slide removal	1200 veh/h
Pavement markers	1100 veh/h
Bridge repair	1350 veh/h

NOTE: Adapted from paper by Dudek and Richards;
Data provided are from California and are expressed as peak hour flow rates.

TABLE 2-3 Capacity Values for Long Term Construction Work Zones

Number of Lanes		Number of	Range of	Average
Normal	Open	Studies	Values	per Lane
			(veh/h/ln)	(veh/h/ln)

3 2 7 1780-2060 1860
2 1 3 - 1550

Source: Dudek, C. L. Notes on Work Zone Capacity and Level of Service. Texas
Transportation Institute, Texas A&M University, College Station, Texas, 1984.

Canadian researchers found (Al-Kaisy et al. 2000) that the average freeway capacity at long term work zones was reasonably close to the corresponding values that were collected by Dudek and Richards (1982) and given in Chapter 6 of the 1997 HCM. They conducted the research on a three lane freeway in which one lane was closed for construction activities. They noted that grades affected freeway capacity at the sites they investigated. Their research also showed that the temporal variations, such as percentages of heavy vehicles and a commuter-dominant driver population, have a significant impact on the freeway capacity at long term construction sites.

WORK ZONE CONGESTION MECHANISMS

A number of concepts are required to explain freeway work zone congestion. The point at which too many cars are trying to squeeze through is the “ bottleneck”. A “ queue” or line of vehicles builds up in advance of the bottleneck. The “ wave” is the speed of the resulting interface or boundary between the queue and free-flowing traffic.

The flow at the onset of queue development on a freeway facility is called the breakdown flow (HCM 2000). When the demand exceeds the capacity, bottlenecks will occur along the highway. This results in backups and delays.

Freeways become congested not only when demand exceeds normal capacity, but also when capacity has been reduced. The closure of one or more lanes due to events ranging from a crash to a work zone lane closure can produce this effect. Reduced capacity makes the demand exceed the sustainable capacity along the section, which will cause the traffic flow to break down. The traffic handling capacity of a work zone is the principal determinant of the magnitude of impacts of a work zone on traffic along a given section of freeway during a given time period and prevailing traffic conditions (Krammes and Lopez 1994).

DATA COLLECTION METHODS

Various methods which have been employed by others to collect traffic flow data at work zones were reviewed. The reviews were confined to the study of methods which were applicable to this research effort.

The review produced references to studies which had been initiated by the state departments of transportation in Nebraska, Iowa, Kansas, Missouri, Minnesota, Texas, and North Carolina. Some of these are discussed in the following sections. Roadway Surface-Mounted Devices

In a work zone on I-80 between Lincoln and Omaha, Nebraska, Nu-Metrics NC-97 traffic counters/classifiers were used to measure the traffic speed and volume data. This unit can measure and record the speed and lengths of vehicles passing over it. These units can be programmed to collect data from a particular time and stop automatically.

Three units were located at 500, 1000 and 1500 feet in advance of the lane closure taper. Of these three units, the one at 1000 feet from lane closure failed to collect data (the reason for failure was not given). In order for these units to work continuously throughout the duration under high traffic volumes, it was not possible for the units to store each vehicle speed measured without exceeding their storage capacity. Therefore, the units were programmed to store the distribution of vehicle speeds during each 15-minute interval. Lidar Guns

Lidar units were used to measure the speed of the vehicles entering the work zone on I-80 in Nebraska. These units are capable of measuring speeds with an accuracy of ± 1.6 km/h ( ± 1 mph). These units were used to determine the efficacy of the Speed Monitoring Displays (SMDs). To avoid cosine error, measurements were taken by placing the units as close to the side of the roadway as possible. Since the density of traffic primarily affects vehicle speeds during congested flows, only the speeds of vehicles with at least five-second headways were measured. Radar Guns

Radar was used to collect speed data as a part of a portable traffic management system (PTMS) operational test on I-94 in Minnesota (SRF 1997). The PTMS included portable machine vision cameras placed at strategic locations in a work zone. These units were used with relative ease in a variety of work zones. This unit also transmitted data to the traffic control center for analysis. The overall system operation was successful. Machine Vision

Machine Vision is another component of the PTMS developed as a part of the Smart Work Zone System for I-94 in Minnesota (SRF 1997). Machine Vision technology uses video cameras and computers to emulate the function of the human eye. It is said that this unit can be deployed in a wide variety of work zones with relative ease. Detection zones placed on the video image emulate an inductive loop and are able to gather data such as speed, volume, occupancy, queue lengths and vehicle classification. Loop Detectors

The traffic volumes were collected from loop detectors on I-94 and I-35 (Minnesota) in the vicinities of two work zones (SRF 1997). The data obtained from these units were compared to the Machine Vision counts as a check. Loop detectors, or sensors in the pavement, relayed freeway speed and volume data to computers.

CHAPTER 3
RESEARCH PROCEDURES

This chapter describes the site selection, equipment selection, data collection, and data reduction. The data needed for the research included the volume, the speed, and the existence of backups for a given work zone site. The volume along the section, presence of an existing AWIS, and other factors influenced the site selection process. To collect these data, suitable equipment had to be identified and collection procedures had to be devised.

COMPARING AWIS WITH OBSERVATIONS

The sites at which the AWIS systems were operating during the data collection effort were along I-40 east of North Little Rock and west of West Memphis. AWIS readings were compared with actual field observations.

The computer files on which the AWIS indication records were stored were obtained from the AHTD. These contained information about what kind of messages had been sent to the changeable message signs (CMS) alongside the freeway, and the time they were sent. The information sent to the CMS reflected readings from sensors located along the work zone. The sensors were spaced at 1.6 km (1 mile) intervals. From this message, it was possible to identify when the AWIS sensed a backup in the work zone. This data was available in the spreadsheet format.

The video tapes made during the data collection process (described in the following sections) enabled the researchers to ascertain when backups occurred in the field. This record from the video tapes was compared with the AWIS log to assess how well the AWIS worked in detecting backups in the work zone.

SITE SELECTION FOR CAPACITY STUDY

The sites at which work zone capacities were studied were confined to rural high-speed freeways normally having two lanes in each direction. Factors considered favorable at a site included:

1. having roadside terrain that provided a suitable location for observers to videotape the traffic;

2. having higher volumes, thus having a higher probability of backups;

3. closure of the outer lane;

4. not having construction work occurring within 0.8 km (½ mile) of the taper beginning; and

5. not being close to a high-volume entry ramp, in order to eliminate impacts of merging traffic.

Since this research was funded by the Arkansas State Highway and Transportation Department, only the Interstate highways in Arkansas were considered for field tests. The Interstates in the state of Arkansas are I-30, I-40, I-55, I-430, I-440, I-530, I-540 and I-630. The section of I-40 between North Little Rock and West Memphis had a considerable amount of reconstruction activity, and the highest volumes in the state (outside of urban areas). All sites selected for study were on I-40 between North Little Rock and West Memphis.

A great deal of coordination among AHTD personnel, MBTC personnel, an equipment rental store, and the contractor was required before field data collection could be scheduled. Telephone calls were made to field personnel to confirm site attributes. Once the sites had been selected, the schedule for gathering data at that site was set.

The first attempt to collect data, on Wednesday, October 25, 2001, did not go well. Although recent traffic volume patterns had been studied in an attempt to determine when backups were likely to form, a new pattern emerged on that day. A backup formed in the late morning, about the time the data collectors arrived to set up the equipment, and the backup remained constant throughout the day. One of the Lidar guns malfunctioned, and had to be returned to the manufacturer for repair. Later in the day, it began to rain. Another attempt which had been coordinated for late November was canceled because inclement weather on the scheduled days interrupted an otherwise long stretch of mild weather.

Data were collected at the following sites in the eastbound (EB) or westbound (WB) direction.

1. I-40 WB Kerr (November 14 and 15, 2001, near the Kerr Road exit)

2. I-40 WB Goodwin (November 16, 2001, between Brinkley and Forrest City)

3. I-40 EB Shearerville (January 9, 2002, west of West Memphis)

4. I-40 WB Galloway (September 4, 2002, east of North Little Rock)

5. I-40 WB County Line (September 5, 2002, at the Pulaski-Lonoke county line)

EQUIPMENT SELECTION

At the first three sites that were studied, three Lidar guns, a Radar gun, and four video cameras were needed to collect field data at four sequential data collection stations along the freeway. Four people were needed at any one time to operate the equipment. Since the number of data collection stations was reduced from four to three at the final two study sites, only three Lidar guns and three camcorders were required. Scaffolding

Scaffolding was set up at some locations to enhance the observer’ s view of traffic. An enhanced line-of-sight was needed at some locations because the roadside area was depressed and there were two parallel lanes of moving traffic (at Stations 1 and 2). In this situation, being elevated improved the ability to collect speed data. Trial and error resulted in the determination that a six-foot high scaffold setup was adequate for the purpose of this study.

Speed Measuring Equipment

Lidar guns were used to measure the speed of the vehicles on the freeway. A Radar gun was also introduced for this purpose because of shortage of Lidar guns. Where there were four data collection stations, the Radar gun was used at one station, while the Lidar guns used at the three other stations. The Radar gun was employed at a station where aiming the device was less critical. The Radar gun batteries were not as reliable as the Lidar guns batteries. Some interruptions in the data can be attributed to Radar gun battery failure.

There are two major advantages of using Lidar guns over Radar gun. First, the Lidar gun can measure distance to a vehicle as well as the speed of that vehicle, while the Radar gun only measures speed. Secondly, the laser beam transmission is much more confined than that of a Radar gun, allowing more precise targeting of an individual vehicle at a distance. Video Recording Equipment

Three 8mm cameras and one VHS camcorder were also used for this research. At the first three sites, 8mm video cameras were used at Stations 1, 2 and 4 while at Station 3, a VHS camera was used. For the last two studies, 8 mm cameras were used at all three stations, while the VHS camera was held as a backup camera. One advantage of the 8 mm camera over the older VHS camera was its longer battery life.

The VHS camera positioned at Station 3 at sites 1, 2, and 3 did not continuously collect data on some occasions, and therefore there were gaps in the data. This was in part caused by the lower reliability of the VHS camera.

FIELD DATA COLLECTION

Speed, volume, and backup data were collected at rural, high speed freeway work zones using the previously described equipment. All data were collected on clear weekdays when the pavement was dry. On the second day at the Shearerville site, there was a very light drizzle, but this data was not used for analysis.

Establishing Stations at Each Site

At the first three study sites (Kerr, Goodwin, and Shearerville), four sequential data collection stations were set up, spaced at 0.8 km (½ mile) intervals. Each station was located at a standard distance from the beginning of the lane closure taper (see Table 3-1). After analyzing the data obtained from the first three data collection efforts, it was concluded that three stations would yield the necessary data for the project. Hence, at the final two sites (Galloway and County Line), Station 2 was omitted. A typical station setup is shown in the following Figure 3-1.

TABLE 3-1 Distances from the Taper Begin to the Stations

Station Position of Station with Respect to Beginning of Taper
Number

1. 1 1.6 km (1.0 mile) upstream
   

2. 2 0.8 km (0.5 mile) upstream
   3 where the taper began
   

3. 4 0.8 km (0.5 mile) downstream
   

Field measurements were made in order to establish the positions of each station. A vehicle odometer was used to measure the distance from the taper beginning to the other stations. The vehicle was stopped on the opposite side of the freeway at the taper beginning and the odometer was set to zero. From there, positions of 0.8 km (½ miles) and 1.6 km (1 mile) in advance of the taper were set. The location of Station 4 was set by driving 0.8 km (½ mile) downstream from the taper beginning.

Figure 3-2 shows the set of stations that was set up at Kerr along I-40 (westbound) in Lonoke County in relation to the mile-markers and AWIS.

Site Idiosyncrasies

Certain aspects of some sites were less than ideal. Based on observation of the site, it was felt that such idiosyncrasies did not adversely affect the quality of the collected data.

At the Kerr site, a low-volume entry ramp joined the freeway immediately in advance of Station

3. At Galloway, a moderate-volume entry ramp joined the freeway at Station 1. At the County Line site, the inside lane was closed, so traffic was shifted first to the inside lane, then to the outside (i.e., “ Iowa weave”). Figure 3-3 presents schematic drawings of the Galloway and County Line setups.

Equipment Setup

After establishing the station locations, the vehicles carrying the scaffolding and the research equipment were brought to the site. The scaffolding was erected at the location and the equipment was set on the scaffolding. The scaffolding was 1.8 m (6 feet) high, allowing an improved line of sight for video cameras and speed-measuring guns.

Scaffolding was used at Stations 1 and 2 at the Kerr site, at Station 2 of the Shearerville site, and at Station 1 at County Line. Station 1 at Galloway was elevated by virtue of being situated on a bridge abutment embankment slope. For Station 3 at the Kerr, Goodwin and Shearerville sites, the cameras and the speed guns were placed inside or next to a vehicle. At all the stations at other stations, the instruments were placed on the ground and scaffolding was not used. Figure 3-4 shows a typical equipment setup upon the scaffolding.

The operators, along with the camcorders and Lidar guns they were using, were not screened or hidden from approaching drivers. But as Figure 3-5 shows, the data collection operation occupied only a small fraction of an approaching driver’ s field-of-view. At one of the stations, the data collector was partially hidden behind the Changeable Message Sign (CMS) assembly (Figure 3-6). Placing the equipment inside or next to a vehicle also concealed the equipment from the approaching vehicle (Figure 3-7).

One person was stationed to collect the data at each station. Since up to four stations were setup at each site, four people were assigned to collect data while one person supervised and addressed problems, such as needing new batteries or video tapes. Thus, the data collection process required that at least five people be present at any given time.

The video cameras were placed next to the speed guns. The times in all video cameras were synchronized before the data collection process commenced. Thus the data from all the stations can be compared in time.

There were some instances when data collection equipment at one or more stations did not operate. This happened because of any of the reasons discussed in “ Equipment Selection”.

DATA REDUCTION

Two types of data were reduced, the AWIS-comparison data and the data for determining at what volumes congestion begins to occur. The videotapes were viewed in the office to glean speeds, volumes, vehicle class, and the presence of congestion. The number of vehicles (or “ distribution” of vehicles) in each lane was also obtained. The data recovered were entered into a spreadsheet and formatted. A set of typical data reduced from the video tapes is shown in Appendix 2. Data for AWIS Comparison

AWIS data were obtained from the AHTD. These files contained information about what kind of messages had been sent to the changeable message signs (CMS) alongside the freeway. The information sent to the CMS was from the sensors located along the work zone. From this message, it was possible to identify when the AWIS sensed a backup. This information was furnished in a spreadsheet.

The field observations of backups were collected on videotapes. The following sections describe the video viewing process. While viewing the videotapes in the office, the status of traffic (backup or no backup) was noted for every minute. Viewing the Video Tapes

Two people reduced the volume data from the videocassettes. One person constantly watched the video and announced what kind of vehicle was approaching and its speed, while the other entered the values in a paper copy of the spreadsheet. This way, care was taken to make sure that no vehicle was missed while transferring data from videotapes to the paper.

A few problems were experienced while viewing the tapes. Whenever a truck occupied the outer lane, it obstructed the data collector’ s view of small vehicles occupying the inside lane. Thus the videotapes had to be viewed several times in order to see a part of or the shadows of the smaller vehicles hidden behind the trucks. At the stations at which there were two moving lanes of traffic and scaffolding was not used, it was sometimes more difficult to identify in which lane the vehicle was traveling. There was probably some error in identifying the accurate lane distribution of vehicles at these stations.

The data were then transferred to the spreadsheet. For the first three sites studied, the volume data were recovered for all four stations. During the subsequent analysis it was found that the volume data retrieved from Stations 2, 3 and 4 were not needed; only the Station 1 volume data were needed for the analysis. Thus it was concluded that volume data from Station 1 would suffice. Hence, for the final two sites, just the Station 1 tapes were examined for volume data. Volume data from other stations were not taken, though the speed data was recovered from tapes from all the stations. Gaps in Data

Occasional interruptions in the data collection process at the various stations resulted in gaps or omissions in the reduced data. At sites 1, 2 and 3, it was decided to consider only those time periods during which the data were available for at least three of the four stations. At sites 4 and 5, data had to be available at all three stations to consider that time period.

Early in the data reduction process, the time and duration of these gaps in the data were noted, producing a record of time periods during which data were not available at all stations. This record helped identify intervals which could be omitted when reducing other data.

COMPARISON OF AWIS DISPLAY WITH OBSERVED BACKUPS

This section discusses the reliability of the AWIS displays to motorists. An AWIS giving drivers information about traffic conditions near or in the work zone was present at some of the study sites. An AWIS was operational in the vicinity of the Kerr and the Galloway sections of I-40.

AHTD downloaded and provided the AWIS computer file records to the researchers for AWIS comparison. The AHTD files showed that the AWIS sensors recorded the backup when vehicles were moving at speeds less than 30 mph. The researchers’ criteria for threshold of a backup was a speed below 50 mph.

The logic function capabilities of the spreadsheet were used to compare field observations with the AWIS record. Table 4-1 shows the sets of logic conditions were developed to compare field observations (“ Field”) of backups with AWIS indications and determine the degree of agreement.

A total of 726 minute-by-minute comparisons of field observations with AWIS-system indications were made. Graphs were prepared to show the agreement of the AWIS displays and the backups observed by the researchers in the field.

The results obtained from the AWIS comparison were graphed using a pie chart (Figure 4-1). The chart showed that the field observations agreed with the AWIS system about 88% of the time.

FORMATTING DATA TO IDENTIFY VOLUMES ASSOCIATED WITH BACKUPS

A number of steps were performed to convert the data collected in the field into a format that could be analyzed with SAS^® statistical software. Other adjustments were made to convert heavy vehicles to passenger car equivalents (PCE).

Large vehicles such as trucks occupy more space and perform more sluggishly than automobiles. Converting trucks and other heavy vehicles to passenger car equivalents allows volume to be compared without a distortion that would be present with varying proportions of trucks in the traffic stream. All data were collected on nearly level and straight freeway sections. Using the conversion factor for “ trucks and other heavy vehicles on a level grade” taken from the Highway Capacity Manual 2000, each truck and other heavy vehicle was converted to 1.5 passenger cars.

ANALYSIS OF DATA TO IDENTIFY VOLUMES ASSOCIATED WITH BACKUPS

A count of vehicles revealed that percentages of heavy vehicles ranged from 45% at the County Line site to 55% for one of the periods at the Shearerville site. The following analyses were performed to identify the volumes at which congestion began to appear. Trellis Plots and Regression Analysis

Preliminary graphs were prepared to view data patterns. When the data were analyzed and graphed, the Trellis plots showed some patterns in the volume-speed-time relationship. The pattern in which these graphs were plotted is shown in Figure 4-2. This graph shows the Station 1 Trellis plot with flow rate (labeled “ volume”) on the x-axis and speed on the y-axis, using time as the conditioned variable in the process (conditioned variable is the one acting as the virtual z-axis in the plot). In order to visualize the data, S-Plus was used to generate several Trellis plots for the Kerr site.

Other graphs were prepared. Figures 4-3 and 4-4 show 1-min traffic volume at Station 1 on the x-axis and speed (mph) at Station 1 on the y-axis; two graphs were needed to show the eight possible combinations of conditions. Within each of the larger graphs, individual plots are placed in various quadrants, based on the absence or presence of backups at Stations 2, 3, and 4. The presence or absence of a backup at Station 1 was indicated with triangular (yes) and circular (no) symbols.

To find a relationship of “ occurrence of backup”as a function of volume, a Robust regression fit was preformed. Figure 4-5 shows this (MM Method) for 1-min volume by speed at Station 1 conditioning on backups at Station 2. The two symbols represent whether or not there was a backup at Station 1. The equations used for Robust Regression are as shown in Equation 4-1, 4-2, and 4-3. These equations were developed by the statistician using the S-PLUS statistical software.

Backup.1= YES, Backup.2= YES, predicted speed1= 6.159585+ 0.7052323(VOL1) Equation 4-1 Backup.1= NO, Backup.2= YES, predicted speed1= 48.32371+ 1.154986(VOL1) Equation 4-2 Backup.1= NO, Backup.2= NO, predicted speed1= 70.42984-0.1040536(VOL1) Equation 4-3

The Trellis plots showed that there was too much randomness in the data. To smooth out these erratic patterns, three to ten-minute moving averages of speed and moving sums of volumes were prepared.

Two types of Trellis plots were designed this time. In the first graph, the 3-minute moving average of speed is on the y-axis and time is on the x-axis, with the 3-minute moving sum of volume as the conditioning variable. In the second graph, volume is on the x-axis and speed is on the y-axis. Here, time is the conditioning variable, as in the previous graph.

The graphs obtained through the Trellis plots and Robust Regression (MM method) did not show any significant relation between the speed-volume-time. The regression analysis did not explain the sudden drop in speed and formation of traffic congestion. This method of analysis did not produce definitive results. Therefore, a different approach was devised to perform the analysis.

Moving Sum-of-Volume Analysis

Since the regression analysis did not explain the sudden drop in speed and formation of traffic congestion, it was decided to create scatter plots of the data with time along the x-axis and equivalent hourly volume (in pce/h) stacked over speed along the y-axis. For an initial trial, three- to ten-minute moving sums of volume from the Kerr Station 1 were calculated.

The first attempt uncovered a need to modify the programming. Missing data produced blank or empty parts in the graph. The software would not consider the intervals during which even one data point (speed data or volume data) was missing. Because of this, the gaps (blank region) kept on widening as the intervals moved on from three to ten minute. This was corrected by programming to take an average of speed of only the available speeds, i.e., if only five speed data readings were available for a six-minute interval, then these five readings were used to calculate the moving average.

The graphs thus obtained gave a clear picture of speed-volume-time relationship. Figure 4-6 shows the graph obtained from the 5-min moving sum of volume and 5-min moving average of speeds.

Though the depiction of backup patterns was improved by these graphs, the graphs for the Shearerville site did not show much difference between backed-up conditions and free flow conditions, because of rounding off of the speed value done by moving average of speed data. Hence, it was decided to prepare a graph which showed both smoothed speed and non-smoothed speed. This kind of graph was prepared only for Shearerville data. Figure 4-7 shows this type of plot. This graph projected the drop in speed much clearer than other plots that had been prepared for Shearerville.

Identifying Backup Periods from the Graph

The data from Goodwin, Shearerville, Galloway, County Line and a part from Kerr (November 14, 2001) were used for this analysis. Data collected at Kerr on November 15, 2001 were not used for analysis. This is because the traffic was constantly backed up much of the time at Kerr, and there were few time periods when congestion disappeared and then reformed.

Once the graphs for all the stations had been prepared, the time periods during which drop in speeds occurred were marked on the graphs. With this as a reference, the spreadsheet was examined and times during which there had been a drop in speed were manually highlighted. A spreadsheet with just the speed data from all the stations and the volume data was prepared for each site. The readings from each station were in adjacent columns, and the entries were aligned so that rows showed speed and volume data from each of the stations during the same minute.

Next, a nearly horizontal mark was extended from the identified backups to the other three stations. The mark was tilted at a slight angle to reflect the speed and travel time from Station 1 to Station 4. In other words, the mark highlighted data across the successive stations and roughly tracked the vehicles in time as they progressed through the stations.

This process identified flow rates at Station 1 that resulted in backups at Station 4. The flow rates that caused the backups were noted in a separate spreadsheet. The equivalent hourly flow rates resulting in backups were counted, by increments of 100 pce/h. For instance, the number of times that a flow rate between 1200 and 1299 pce/h resulted in a backup were counted. Also counted were the number of times that this flow rate was not accompanied by a backup.

From this, the percentage of time that backups occurred in each flow rate range was calculated. Different graphs were plotted for each moving sum flow rate (1-min, 2-min, 3-min and 5-min). The graphs were plotted using a third degree polynomial equation. Different combinations of parameters were removed from the third degree equation and the results plotted to get the best fit. The coefficient

2

of determination (R ) value was also calculated for each of these plots and the best plots were found.

FLOW RATES ASSOCIATED WITH BACKUPS

The final data sheet had columns for the equivalent hourly flow rate (in passenger car equivalents -- pce) range, number of times a backup occurred, number of time a backup did not occur, and percentage of time a backup occurred. Figures 4-8, 4-9, and 4-10 show plots with probability of backup (i.e., percentage of time that a backup occurred) on the y-axis. The x-axis is the short-term flow rate expressed as an equivalent hourly volume. The equations (Eq. 4-4. 4-5, 4-6) associated with each graph follow. The equations obtained from the one-minute, three-minute, and five-minute moving sum flow rates had high R² values. But their higher R² values indicate that the predictive powers of the three- and five-minute interval equations were clearly superior to those of the one-minute equation.

1-Minute flow rates ranging from 750 to 1550 pce/h y = 5.048E-10 X³ + 0.000 000 599 3 X² + 0.119 (R² = 0.82) Eq. 4-4

3-Minute flow rates ranging from 750 to 1350 pce/h y = 3.585E-9 X³ - 0.000 008 86 X² + 0.007 31 X - 1.996 (R² = 0.98) Eq. 4-5

5-Minute flow rates ranging from 750 to 1250 pce/h y = 9.463E-9 X³ - 0.000 025 78 X² + 0.0243 X - 7.053 (R² = 0.97) Eq. 4-6

The graphs plotted for 2-min moving sum volumes were excluded from the results because all the graphs obtained from that data were curving upwards at the left end after reaching the nadir. This would have suggested a higher probability of a backup as the volume decreased.

The one-minute flow interval equation, in the range below 900 pce/h, gives rounded probabilities of 0. For flow less than 1000 pce/h, the observed probabilities were 0.

CHAPTER 5
CONCLUSIONS

Motorists traveling through work zones are faced with both routine and unexpected tasks of driving coupled with additional distractions provided by the work zone. This has created safety problems in work zones despite significant applications of work zone traffic control devices. Many work zones require the closure of one or more through-traffic lanes. When traffic volumes exceed the capacity of these merge points or of the reduced number of lanes, the resulting congestion can lead to the formation of queues, which result in delays and increase the potential for traffic crashes. Transportation agencies sometimes implement an Automated Work Zone Information System (AWIS) at work zones to notify approaching motorists that they face a slowdown and a backup ahead.

This research was conducted at work zones on rural freeways at which two lanes of traffic in a direction were reduced to one lane. The objectives were to compare the displays of an AWIS with field observations of traffic conditions, and to determine at what volume backups were forming in these work zones.

PREVIOUS RESEARCH

The congestion, delay, and other problems associated with work zones led many researchers to investigate the capacity values related with work zones and to improve the safety of the work zones. Other studies have estimated work zone capacity values. They suggested that the capacity values differ with the location of the work zones. The type of barriers used, grade, presence of diverging, merging or weaving, alignment, trucks, environment, and other factors affect the capacity values for the work zones. Reported per-lane capacities range from 1100 veh/h to over 2000 veh/h.

RESEARCH PROCEDURE

The computer files storing the AWIS records were obtained. The comparison of the AWIS indications with field observations indicated the AWIS gave correct indications about 88% of the time.

Data to identify volumes at which backups occurred were collected at five different locations along I-40 between North Little Rock and West Memphis. At these five locations, studies were conducted on six different days during different times of the year. Data were collected with video cameras and Lidar/Radar guns, placed at 3 or 4 sequential stations at a site.

Preliminary analyses using the Trellis Plot and Robust Regression (MM method) were performed on the data. A Moving-Sum-of-Volume analysis was done on the data to explain the sudden drop in speed and formation of traffic congestion. Graphs were created to show the relationship between time, volume and speed. These graphs revealed that during most of the peak volume periods, there was some decrease in speed. These processes did not help define the volume associated with the formation of congestion. However, inspection of the outcomes helped identify an alternative approach.

In the subsequent analysis, the peak volumes associated with each speed drop were entered into a separate spreadsheet. From this data, the number of times backups occurred and the number of times backups did not occur were counted for a each one-direction flow rate range of 100 pce/h (on these straight and level freeways, heavy vehicles were counted as 1.5 passenger car equivalents). The percentage of time backups occurred for each flow-rate range and these results were plotted using a degree three polynomial equation. The associated equations presented in the previous chapter predicted the probability of a backup at a given one-direction flow rate.

The equations can be applied in the following manner. Using the three-minute equation, solving for x = 1250 pce/h, produces a result of 0.30. This means that, based on findings from the field data, there is a 30% probability of backup if the one-direction flow rate for three minutes is equivalent to 1250 pce/h (62.5 passenger car equivalents in three minutes). If it were to last for a five minute duration, this same flow rate has a 44% probability of occurrence.

OBSERVATIONS

Although not one of the research objectives, factors contributing to rural freeway work zone backups were observed during the course of during the data collection process.

1. All observed backups began within the one-lane section, and were propagated upstream into the two-lane section. No instances of a backup resulting from congestion at the merge point were observed. It should be noted that the advance warning and regulatory signs required motorists to merge in advance of the actual taper, and there seemed to be a high degree of compliance with this requirement.

2. Once a backup forms, it can be hard to eliminate. Backups tend to persist.

1. Contractor activities can contribute to the formation of backups. Such activities observed included obstructing traffic lanes while repositioning equipment, and construction vehicles slowing in the remaining single lane before exiting into the work zone space. The latter was observed where there was an adjacent frontage road which perhaps could have been used to

2. route construction traffic on.
3. Short taper ramps can contribute to the formation of backups. There was a taper exit ramp to another freeway downstream of Station 4 at Galloway. Some, but certainly not all drivers, seemed to slow down unnecessarily to take this exit, thus backing up traffic behind them. Perhaps the presence of a parallel exit ramp would have allowed the drivers to leave the through lane before they began to slow down, thus eliminating the source of a backup.

CONCLUSIONS

The findings from this study of work zones on rural freeways at which two lanes of traffic in a direction were reduced to one lane are summarized below. The study sites were fairly straight and level. Fractions of heavy vehicles in the traffic stream ranged from 45% to 55%.

1. The AWIS installed at the work zones concurred with the manual observations at the study sites about 88% of the time.

2. The probability of a backup occurring increased with an increase in volume.

3. Backups formed at short term flow-rates of as low as 800 to 900 pce/h (passenger car equivalents per hour).

4. At a one- to five-minute flow rate of up to approximately 1100 pce/h, the probability of a backup occurring is 10% or less. While this may sound low, the implication is that 1 time out of 10 when a short term one-direction flow rate is less than 1100 pce/h, a backup will form.

Backups were observed to occasionally result from short-term flow rates that were lower than capacities reported in the literature. Two possible explanations follow.

1. In rural freeway work zones of the type studied, with high percentages of heavy vehicles, capacities are occasionally less than the values found in the literature.

2. Given the high percentage of heavy vehicles on the freeways studied, the conversion of one heavy vehicle to 1.5 passenger car equivalents (as listed in the 2000 Highway Capacity Manual, page 23-10) inadequately accounts for the effects of heavy vehicles. Perhaps a greater factor, such as 2.0, would better represent the effects of heavy vehicles in work zones similar to the ones studied, but no such conclusion can be made from this study.

blank page

REFERENCES

Al-Kaisy, A., Zhou, M., and Hall, F. (2000) “ New Insights into Freeway Capacity at Work Zones: Empirical Case Study.” Transportation Research Record 1710. Transportation Research Board, Washington, DC.

Al-Kaisy, A, and Hall, F. (2001) “ Examination of Effect of Driver Population at Freeway Reconstruction Zones.” Transportation Research Record 1776. Transportation Research Board, Washington, DC.

Dudek, C. L. and Richards, S. H. (1982) “ Traffic Capacity Through Urban Freeway Work Zones in Texas.” Transportation Research Record 869. Transportation Research Board, Washington, DC.

Krammes, R. A., and Lopez, G. O. (1994) “ Updated Capacity Values for Short-Term Freeway Work Zone Lane Closures.” Transportation Research Record 1442. Transportation Research Board, Washington, DC.

Pesti, G., and McCoy, P. T. (2001) “ Long-Term Effectiveness of Speed Monitoring Displays in Work Zones on Rural Interstate Highways.” Transportation Research Record 1754. Transportation Research Board, Washington, DC.

Polus, A., Craus, J., and Livneh, M. (1992) “ Flow and Capacity Characteristics on Two-Lane Rural Highways.” Transportation Research Record 1320. Transportation Research Board, Washington, DC.

SRF Consulting Group, Inc. (May 1997) “ Portable Traffic Management System - Smart Work Zone Application - Operational Test Evaluation Report.” Prepared for MnDOT. SRF No: 0942089.7/11

Transportation Research Board. (1997) Highway Capacity Manual. Washington, D.C.

Transportation Research Board. (2000) Highway Capacity Manual. Washington, D.C.

Tudor, L.H., Meadors, A., and Plant, Robert II. (2003) “ Deployment of Smart Work Zone Technology in Arkansas”. Transportation Research Board 2003 annual meeting paper 03-3115.

blank page

APPENDIX 1

Typical Speed Data Reduced from Lidar Gun

APPENDIX 2

Typical Volume Data Reduced from Videotape