Generating Performance Measures From ... - Semantic Scholar

4 downloads 16 Views 945KB Size Report
Email: [email protected] ... Performance measures generated from archived ITS data can be ..... An automated procedure to generate such a graphic.

Generating Performance Measures From Portland’s Archived Advanced Traffic Management System Data

Robert L. Bertini Department of Civil and Environmental Engineering Portland State University P.O. Box 751 Portland, OR 97207-0751 Phone: 503-725-4249 Fax: 503-725-5950 Email: [email protected] Monica Leal Department of Civil and Environmental Engineering Portland State University P.O. Box 751 Portland, OR 97207-0751 Phone: 503-725-4297 Fax: 503-725-5950 Email: [email protected] David J. Lovell Department of Civil and Environmental Engineering University of Maryland 1179 Glenn Martin Hall College Park, MD 20742 Phone: 301-405-7995 Fax: 301-405-2585 Email: [email protected]

March 2002

Bertini, Leal and Lovell


Generating Performance Measures From Portland’s Archived Advanced Traffic Management System Data Robert L. Bertini, Monica Leal and David J. Lovell ABSTRACT Certain performance measures were generated for a freeway corridor in Portland, Oregon (eastbound US 26) using archived loop detector data. The US 26 Sunset Highway is a major east-west corridor connecting downtown Portland to the burgeoning west side, including major residential communities as well as Silicon Forest, containing the region’s high tech industry. The study shows that with the use of real data, it is possible to determine the functionality of the facility with respect to measures such as mobility, economic development, quality of life, the environment, resource conservation and safety. Because surveillance systems are often already in place for traffic management purposes in urban areas, archived data can easily be used to develop performance measures in real time and to track them over time. This concept can be applied to specific corridors of interest or to an entire metropolitan area.

This kind of

information can help agencies have a better vision of the current performance of the transportation network, its evolution over time, as well as aiding in setting clear goals and objectives to improve the performance of the facility. These data may also be used to calibrate and/or validate travel demand forecasting models.

Bertini, Leal and Lovell


INTRODUCTION With the implementation of numerous Advanced Traffic Management Systems (ATMS) as part of our nation’s Intelligent Transportation Systems (ITS), many jurisdictions are realizing that the data collection and surveillance systems used to operate transportation systems on a day-to-day basis can be used as rich sources of data that can be useful for many purposes. Once this vast quantity of data is archived, processed and converted to useful information, it is possible to generate performance measures to aid in the planning, design and operation of transportation systems. There is a nationwide movement toward development of performance measures for operations, as evidenced by the National Dialogue on Operations led by the U.S. Department of Transportation. The vision adopted by the National Dialogue is “managing and operating the existing transportation system so that its performance meets or exceeds customer expectations.” Another indicator of the movement toward the greater use of performance measures is the recent publication of the National Cooperative Highway Research Program (NCHRP) Project 8-32(2), Multimodal Transportation: Development of a Performance-Based Planning Process. (NCHRP 1999) Also, the newest generation of travel demand forecasting, such as Metroscope and TRANSIMS (currently being developed in Portland), require vast amounts of performance data for validation and calibration. Performance measures generated from archived ITS data can be valuable inputs to these models. As one example of the ongoing development of performance measures on a statewide basis, Table 1 shows a menu of preliminary recommended performance measures (safety and mobility categories) being considered for adoption by the Oregon Department of Transportation (ODOT). By definition, performance measures generated on a statewide basis must be simple enough to

Bertini, Leal and Lovell


satisfy the “least common denominator” in terms of data availability and level of detail. Often statewide measures can seem somewhat watered-down, since rural areas are usually not rich in surveillance capabilities. Clearly, in urban areas with highly developed ATMS it is possible to provide more detailed measurements on a corridor and/or regional level. With this recognition, the objectives of this paper are to: Review current literature documenting performance measures of interest Experiment with real archived ATMS data to expand and/or focus statewide level measures to a corridor level Demonstrate what may be possible to implement on a system-wide basis for tracking freeway performance. Consider possible corridor level performance measures. Next, we will describe the test bed chosen for this experiment, as well as the sensor data that were used. In the following section, the NCHRP Project 8-32(2) guidelines are used to generate various sample performance measures in categories of mobility, economic development, quality of life, the environment, and resource conservation and safety. In addition, some data fusion combining incident data, automatic vehicle location (AVL) data from incident response vehicles, and loop detector data are presented. Finally, the manuscript ends with some observations and conclusions.

DATA ODOT, the City of Portland, Tri-Met (Portland’s transit agency), Metro (regional government), and other regional jurisdictions have developed the TransPort (Transportation Portland) program. This organization brings together the various agencies to develop, operate and evaluate

Bertini, Leal and Lovell


traffic management, incident response, and traveler information systems as part of the region’s ATMS. The program’s goals are to reduce congestion, reduce travel times, and minimize and prevent accidents on the transportation network in the Portland region (including Vancouver, Washington). Portland’s ATMS includes freeway ramp meters at nearly every on-ramp. For this experiment an 11-mile corridor along eastbound US 26 between Helvetia Road and Skyline Road was chosen as shown in Figure 1. The eastbound freeway has two lanes between stations 1 and 9 and three lanes at station 10. This freeway is the major east-west corridor connecting downtown Portland to the burgeoning west side, including major residential communities as well as Silicon Forest, containing the region’s high tech industry. The traffic surveillance system in this corridor, installed as part of the implementation of the ramp metering system, consists of 10 mainline inductive loop detector stations (with pairs of detectors located in each lane) and associated onramp detectors at 10 eastbound on-ramps. The detector stations are labeled 1 through 10 (ramp volume data at station 8 includes volumes from OR 217 plus Parkway Road, which merge upstream of the ramp meter).

The data recorded by these sensors include vehicle count,

occupancy (percent time that the detectors are occupied by a vehicle) and average speed as measured by the detectors in each lane and on each on-ramp and are aggregated locally every 20 seconds. These 20-second data are transmitted to the traffic operations center (TOC) via the regional fiber optics network. Typically ODOT archives their data at a 15-minute level; but for this research, data were archived in the most raw form available (20 seconds). The time period chosen for this experiment was the week of Monday, October 30 to Friday, November 3, 2000. The presence of congested conditions in this corridor heightens the need for estimating the quality of its performance. Using archived data from loop detectors, performance values can be

Bertini, Leal and Lovell


calculated not only to assess safety and mobility characteristics now, but also to have the ability to track these indicators over time.

DATA ANALYSIS The NCHRP 8-32(2) guidelines include a library of performance measures proposed for adoption by transportation operating agencies, so that decisions can be based on actual performance measurement. The measures considered are divided into categories of: mobility, economic development, quality of life, environmental/resource conservation and safety. Calculations of sample measures using Portland’s archived ATMS data are included in the following subsections. It is clear that there are numerous ways to visually display performance data, but the NCHRP guidelines do not focus on this important area of research. While some efforts are made here to display performance data in unique ways, this paper is by no means comprehensive. For further explorations of helpful means of displaying data, see several papers produced by the Washington State Transportation Center (Nee et al. 2000, Ishimaru et al. 2001, Ishimaru and Hallenbeck 1999).

Mobility Mobility measures provide indications of how easy or difficult travel can be along a corridor or in a region. Performance measures related to the mobility of passengers or freight are defined in this section. Suggested mobility measures according to NCHRP 8-32(2) include congestion measures, such as delay, volume-to-capacity (V/C) ratio, level of service (LOS), trip time, amount of travel such as vehicle miles traveled (VMT) and vehicle hours traveled (VHT), mode

Bertini, Leal and Lovell

share, transfer time, and transit performance.


The following performance measures were

developed as indicators of the quality of mobility of the corridor:

Average Daily Traffic (ADT) Figure 2 shows the ADT measured during the weekdays of October 30 to November 3. The ADT values were calculated by obtaining the average weekday 24-hour volume at each detector. Also shown in the figure are the ADT data from the ODOT permanent count recorders (PCRs) and tabulated statewide by the ODOT Transportation Systems Monitoring Unit (ODOT 1999). The corridor’s bi-directional ADT reported by the PCRs is also shown in Figure 2. In order to compare the ADTs, a directional distribution of 50/50 is shown (westbound ATMS data were not provided). As shown in the figure, the ADT generated from the archived loop detector data is very close to the PCR measures when a directional distribution of 50/50 is assumed, until station 8. At station 9 the directional distribution is reduced for the eastbound direction and at station 10 the directional distributional is close to 50/50 again. This is important because ODOT could replace their PCR data with the ATMS data and save data collection costs.

Average Daily Traffic Per Freeway Lane ADT per freeway lane was also calculated for weekdays during the study period (not shown). This analysis confirmed that the left lane (lane 1) had higher flows than the other lanes except at stations 1 and 7, where shoulder lane (lane 2) flows were higher. However, the difference between volumes in the individual lanes at these stations was very small. At station 1 the difference was nearly 1000 vehicles, and at station 7 the difference was about 350 vehicles.

Bertini, Leal and Lovell


Also, at station 8 the ramp volume was higher than the volume in lane 2, where there is a heavy movement from OR 217.

Average Speed Figure 3 displays the average speed for the corridor on one day (November 1), where the raw 20sec speed data are plotted using the right-hand axis. The speed reductions can be observed during the a.m. and p.m. peak periods. Figure 3 also shows the speed plotted cumulatively (lefthand axis), where the slope of this curve represents the speed (normalized by the 20-sec time intervals). To magnify the curve’s features, it has been skewed (consider that the difference between the raw cumulative curve and a line V=v0t´ has been plotted, where v0 is the skew slope and t´ is the elapsed time from the beginning of the curve). The dashed lines represent a linear approximation of the speed (estimated by eye). At 7:21:20 a.m. the speed dropped from 60 to 49 mph, followed by an increase to 58 mph. At 2:55:20 p.m. the speed dropped to 49 mph, but at 4:01:20 p.m. the speed dropped further to 36 mph. This speed continued until 7:36:20 p.m. when the speed increased to 59 mph. It is clear that plotting the speed data cumulatively reveals changes in freeway conditions in a remarkably clear manner.

Travel Time Derived from speed, but of more relevance to the user (in practice and as modeled in travel demand forecasting systems), the total corridor travel time was calculated by summing the estimated travel time for each station. For estimation purposes, the influence area of each detector was assumed to be the distance between the upstream and downstream midpoints between each detector pair. The travel time was obtained by dividing the length of each

Bertini, Leal and Lovell


influence area by the speed measured at the respective detector. The (raw) total travel time was plotted versus time for November 1 in Figure 4. Increases in travel time during the a.m. and p.m. peak periods can be observed, as well as the free flow travel time of about 11.7 min. This value was estimated using data during the off-peak periods (overnight) when vehicles travel unimpeded. Figure 4 also shows the travel time plotted cumulatively (the slope of the curve represents the travel time normalized by the 20-sec time interval), which allows an observer or traffic manager to see clearly when travel time is changing by tracking the slope of the curve as it deviates from the free flow travel time curve. An increase in travel time is observed at 7:21:20 a.m. (11.7 to 17 min.) following by a decrease at 8:16:20 a.m. (17 to 11.7 min.). The travel time increased again at 2:55:20 p.m. (11.7 to 25.9 min.) until 7:56:20 p.m. (25.9 to 11.7 min.) when the travel time decreased. The two periods when the travel time increased were identified as the peak periods of the day being studied.

Vehicle Miles Traveled The total VMT was estimated for the study corridor (on weekdays). We assumed that the vehicles counted on the mainline traveled those entire segments and that the vehicles using the ramps traveled over half of each segment. The length of each section represented by each detector was multiplied by the count measured at each detector to obtain the VMT, which was then summed to arrive at the total. Table 2 shows the results of this calculation.

Person Miles Traveled The PMT was also calculated for the study corridor by multiplying the VMT by the average vehicle occupancy, depending on the composition of the traffic. The traffic composition,

Bertini, Leal and Lovell


comprised of the percentages of each type of vehicle traveling through the corridor, was obtained from ODOT and the following average vehicle occupancies were assumed: 1.42 for passenger cars (source: ODOT), 1.0 for motorcycles, 25 for buses, and 1.1 for any kind of truck. The final PMT for the whole corridor for the weekdays can be observed in Table 2.

Mobility Index The measures presented thus far have been derived unambiguously from real archived data. Some performance indices are composites, combining a variety of measures into a single value. This is often done to reduce the complexity and volume of the performance measures and to compare the performance of different facilities and among different modes. As one example, a Mobility Index was generated by dividing PMT by VMT and multiplying by average speed. The Mobility Index, a multimodal index, is shown in Table 2. The Mobility Index used in this paper is proposed by NCHRP 8-32(2) (NCHRP 1999), and reflects a weighted speed where the weighting coefficients are the vehicle occupancies. In regions such as Portland, multimodal considerations are becoming increasingly important.

Vehicle Hours Traveled The total VHT was calculated for the study corridor (on weekdays). The mainline volumes and ramp volumes were multiplied by the travel times (using the process described above) to obtain the total time spent in the system. The total VHT corresponds to the sum of the VHT at every station. Table 2 shows the results of this calculation.

Bertini, Leal and Lovell


Person Hours Traveled The PHT was calculated for the study corridor by multiplying the VHT by the average vehicle occupancy as above. The final PHT for the corridor on weekdays can be observed in Table 2.

Vehicle Miles Traveled By Congestion Level Occupancy is directly measurable, and can be used as an indicator of congestion level. Occupancy is the percent of time that a loop detector is occupied by a vehicle. Table 2 shows the standards of occupancy that are used to classify the level of congestion. They are uncongested, near-capacity, and congested flow conditions (May 1990). Table 2 shows the VMT by level of congestion of the corridor for weekdays.

Person Miles Traveled By Congestion Level Use the same process as above, Table 3 shows the PMT by congestion level for the study corridor.

Percent of VMT at a Particular Level Of Service Since freedom to maneuver within traffic and proximity to other vehicles are issues of concern, different levels of service are often used to reflect these measures. As an example, the percent of VMT at a particular LOS (based on the V/C ratio) was calculated for November 1, as shown in Table 3. The highest proportion of VMT was for LOS D (39%) where speed began to decline and density began to increase. Of the VMT, 21% occurred at LOS E and F.

Bertini, Leal and Lovell


Percent of The Freeway Uncongested During Peak Hours If we assume that the congested periods occur when the occupancy is higher than 28%, we can calculate the percent of the highway that is uncongested during the peak hours. This analysis (for November 1) showed that more congestion occurred between 5:30 to 6:30 p.m., where only 54% (6 miles) of the freeway corridor was uncongested.

Number And Percent Of Lane-Miles Congested The total number of lane-miles that were considered corresponds to the length of the segment multiplied by the number of lanes, for a total of 22.4 lane-miles. Figure 5 outlines the peak periods with high degrees of congestion for November 1. The highest value (85%) was observed at 6:48:40 p.m., corresponding to a total of 19 congested lane-miles in the corridor.

Lost Time Due To Congestion Figure 6 shows the time lost due to congestion for November 1. The lost time is the difference between actual travel time and free-flow travel time. Sometimes the lost time due to congestion can be nearly 1 hour, which leads to high delay costs for passengers and freight.

Demand Vs. Capacity Demand was considered in relation to capacity (assumed to be 2,000 veh/hr/lane) for November 1 as an example. Using data from stations 1 through 9 (the two-lane section), Figure 7 shows the demand versus time where one can observe that the only station that exhibits demand that exceeds the assumed capacity is station 7 during the a.m. peak; during most of the day the demand remains close to the assumed capacity. An even more useful tool would be to construct

Bertini, Leal and Lovell


an electronic superimposition of many days’ worth of these plots and to develop a capacity profile by looking at the upper envelope. An automated procedure to generate such a graphic could help in the daily activities of the region’s traffic managers and planners. The V/C ratio is also shown on the right-hand axis. As shown, using this day’s data, stations 1 to 4 are operating at LOS B and C from 6:00 a.m. to 8:00 p.m. whereas stations 5 to 10 are between D and E. The only station that presents LOS F is station 7 during the a.m. peak.

Percent of VMT Which Occurs On Facilities With Particular V/C Ratio While this performance measure may be more meaningful for comparing one facility to another, we can calculate the percent of VMT (by station) that is greater than V/C=0.68 as an example. Table 3 shows the results calculated for November 1. The station with the largest percentage was station 6, with 84%. In addition, it is worth noting that stations 5 to 9 all exhibit more than 70% of VMT at V/C more than 0.68.

Delay Per Vehicle Miles Traveled The delay was calculated for the freeway lanes (not including the ramps) for November 1. In order to obtain the delay the following operations were performed by segment of the corridor and by time of day: Pace (minutes/mile) = 60/Average Speed Delay Rate (minutes/mile) = Pace − (60/Free Flow Speed) Total Delay (veh-hr) = Delay Rate × VMT/60

Bertini, Leal and Lovell


Having the delay and the total VMT for each time period, the delay per VMT is the total delay divided by the VMT. Table 3 shows the results of these calculations for November 1. The highest delay per VMT is observed at station 4, with a value of 1.05 minutes/mile. The total delay per VMT for the entire corridor is 4 minutes/mile.

Reserve Capacity The reserve capacity is the total VMT “capacity” minus the VMT for the entire day. As an example, the reserve capacity was calculated for November 1 for every station (not shown). Between 6:00 a.m. to 8:00 p.m. the difference between the VMT capacity and the actual capacity was very low particularly for stations 7, 9, and 10. At these stations the corridor was operating very close to capacity, especially during the a.m. peak.

Economic Development, Quality Of Life, Environmental And Resource Conservation In this paper some performance measures are calculated in order to gain an understanding of the economic impacts of congestion on eastbound US 26. Using an automated procedure, traffic managers could receive a running tab of costs, using some of the assumptions described below.

Cost of Delay Figure 4 shows the cumulative travel time curve for November 1. Delay is the difference between actual travel time and free flow travel time, so the total vehicular delay is just the area between the two curves in the figure. The cost of delay can be estimated by multiplying an average value of time ($17.87 per person-hour in Oregon, according to ODOT) by the total delay in person-hours. The total delay for November 1 was estimated to be about 2,950 veh-hr, or

Bertini, Leal and Lovell


4,390 person-hours. Thus the cost of lost time for this corridor was about $78,000 for one day. Assuming that a year has 250 working days and that this delay is close to the average, the annual cost of lost time would be over $19 million. The impacts on the movement of freight could also be considered in a similar fashion.

Fuel Cost Fuel costs were approximated using the following simple General Motors model (Daganzo and Newell 1995): E = k5 * L + k 6 * T


E = additional fuel consumed per vehicle k5 = 90 ml/km = 0.038 gallons per mile L = distance traveled in queue k6 = 0.44 ml/sec = 0.418 gallons per hour T = travel time in queue This model estimates the additional fuel consumed by vehicles moving slowly in traffic (the average speed for the a.m. and p.m. peak was 39 mph), such as in a queue. For this estimate, it was assumed that the cost of fuel (in November 2000) was $1.65 per gallon.


Equation 1 can be simplified to: E = $3.15 per hour per vehicle


Thus, the total additional fuel cost due to delay for the corridor on November 1 was 3.15 times the total veh-hrs of delay, or $9,300. On an annual basis this might equate to about $2.3 million. Thus the total cost due to congestion on eastbound US 26 might be about $21.3 million.

Bertini, Leal and Lovell


Safety Every agency wants to maintain a high degree of mobility on the facilities it is responsible for, while keeping them safe as well. Accidents, breakdowns, and other kinds of incidents have an adverse economic impact on society. Portland’s ATMS includes an incident management program with a computer-aided dispatch system (CAD), which archives all incident data. We tested the fusion of incident data with loop detector data to determine whether there are benefits to be gained from combining data from multiple sources.

As an example of the impact of non-recurring congestion on US 26, a major accident identified through the CAD system was analyzed. In order to understand the impact of the accident on congestion, two speed contour maps were generated displaying the data for the same location as where the accident occurred, but on different days. The accident occurred on October 31 on eastbound US 26, between stations 9 and 10 (Figure 1). The confirmed time of the accident was 6:18 p.m., and the estimated end time of the accident was 7:06 p.m.

In order to identify the temporal and spatial extents of congestion, speed contour plots were used (occupancy contour plots can also be used). Figure 8 shows speed contour plots for the US 26 corridor on October 31 and November 1. The x-axis represents the time of the day (24 hours), the y-axis represents the stations, and the color variation represents speed.

Figure 8 shows that the speed drops at the accident location on October 31, and displays the speeds at the same location on November 1. The speed was reduced to a range of 0 to 10 mph

Bertini, Leal and Lovell


due to the accident, when on a different (normal) day the speed at the same location and at the same hour was between 30 and 50 mph.

Fusion Of Incident Response AVL Data With Loop Data The Portland ATMS includes an incident response program (COMET) that includes motorist assistance trucks equipped with AVL equipment. In this section we explore the fusion of AVL data with loop detector data. Figure 9 shows the trajectory (in the time-space plane, where the slope of the trajectory at a point is the speed of the vehicle at that point) of one COMET vehicle and a hypothetical trajectory of a vehicle according to the loop detector speed data. The dashed line represents a linear approximation of the vehicle’s speed, estimated by eye. This COMET trip indicates that the truck’s average speed was about 52 mph between stations 1 and 9 (the loop detector speeds would predict an average speed of about 59 mph). Next, the truck stopped for a few seconds, starting again with an average speed of 14.61 mph, finally reaching a speed of 64 mph, and finished the trip through the study corridor. The travel time for the COMET vehicle along US 26 was 16 minutes, while for the other vehicles it was about 11 minutes. The fusion of the loop data with the AVL data should be explored in the future, but as is clear here, the COMET vehicles are traveling more slowly, which is consistent with their mission to observe the speed limit and be on the lookout for incidents.

CONCLUSIONS Based on a review of recent literature, some valuable performance measures were generated for one freeway corridor in the Portland metropolitan area. This experiment has shown that using archived loop detector data, it is indeed possible to obtain information assessing the functionality

Bertini, Leal and Lovell


of the facility. In accordance with the data available, a limited number of sample performance measures were chosen in this paper.

The information can assist in determining the best

performance measures for obtaining simple, quick and accurate information describing the functionality of the corridor, enhancing decision-making in both real time and over longer planning horizons. With simple, directly measurable variables such as vehicle count, speed, occupancy and incident information, it is possible to set some standards and determine some performance measures that can be generated for the transportation systems around Portland and in that way keep track of the general performance of the network.

Additional information could help generate additional performance measures, such as aerial photography, video surveillance, license plate matching, floating car/probe vehicle information, and cellular phone location data, to name a few. The US 26 corridor is a multimodal one with a parallel light rail transit line and extensive bus service. Performance measures relating to other modes should be considered as well.

Further, spatial data such as density, income and

population can be combined with transportation data to obtain some interesting performance measures. Some examples include the percent of population that can access services with a given travel time and a given speed and the average travel time for different employment centers.

ODOT has made solid decisions to implement an ATMS system and has embarked upon a statewide process to design a performance measurement and evaluation system. Now, it is important to consider using the ATMS system to facilitate the development of performance measures for the Portland freeway system, and more broadly, for the entire transportation network. As a starting point, it would be useful to automate the generation of many of the

Bertini, Leal and Lovell


performance measures and graphs presented in this paper. This automation could serve several useful purposes. For example, without interfering with current management systems, we envision a stand-alone computer terminal that will display all of these graphs and performance measures as it extracts the data directly off the data feed. By putting this information in the hands of ODOT traffic managers, incident responders, planners, etc., for their daily activities, it will be possible to observe which ones get used for which purposes. This system could collect suggestions for improvement and ultimately be integrated into ODOT’s standard operating procedures and computer systems. This somewhat modest approach allows for the possibility that researchers cannot predict a priori which elements of a largely human intelligence-based decision mechanism can best be supported by automated data.

As an integral part of a performance measurement generator, appropriate trends can be tracked over weeks, months, and years and graphical presentations and summary tables can be constructed automatically. Examples include averages and outer envelopes of characteristics such as the physical extent of congestion from the occupancy plots like Figure 8, and a capacity profile from plots like Figure 7. These data can be incorporated into the new data-intensive activity-based planning models (such as Metroscope and TRANSIMS), avoiding costly independent data collection efforts. The traffic surveillance system allows for the monitoring of queue location, which is valuable information for use in calibration and validation of simulationbased planning and operational models.

With regard to capacity, it would be necessary to augment the graphical presentation of an outer capacity envelope with some other intelligence to indicate the degree to which we are confident

Bertini, Leal and Lovell


that we have observed capacity conditions. It is possible that the outer flow envelope could just represent the highest flows seen to date, but with more capacity in reserve. As an example, our confidence improves where we have seen combinations of high flows and high densities.

The value of such an ongoing generator of performance data is that it eliminates the need to make assumptions/estimates about time-varying behavior that find their way into aggregate performance metrics. With our proposed system, the data are being archived every day, so these averages can be calculated, rather than estimated. Finally, in cooperation with ODOT, it would be possible to make some of the performance information available via the Internet. Using an automated procedure, we could observe hit rates usage statistics. Finally, for research purposes, an automated means of soliciting customer suggestions would be included.

ACKNOWLEDEMENTS Dennis Mitchell and Jack Marchant of the Oregon Department of Transportation Region 1 for generously provided the data used herein. Andrew Tang of Cambridge Systematics provided access to the NCHRP materials and the Departments of Civil and Environmental Engineering at Portland State University and the University of Maryland provided funding for this work.

Bertini, Leal and Lovell


REFERENCES Daganzo, C.F. and Newell, G.F. 1995. Methods of Analysis for Transportation Operations. Berkeley, CA: University of California, Institute of Transportation Studies. Garrison, W.L. and J.D. Ward. 2000. Tomorrow’s Transportation: Changed Cities, Economies, and Lives. Boston, MA: Artech House. Ishimaru, J. M., and M. E. Hallenbeck. 1999. FLOW Evaluation Design Technical Report. Seattle, WA: Washington State Transportation Center. Ishimaru, J., M. E. Hallenbeck, and J. Nee. 2001. Central Puget Sound Freeway Network Usage and Performance. 1999 Update. Seattle, WA: Washington State Transportation Center. May, A.D. 1990. Traffic Flow Fundamentals. New Jersey: Prentice Hall. National Cooperative Highway Research Program (NCHRP). 1999. Multimodal Transportation: Development of a Performance-Based Planning Process, Project 8-32(2). Washington, DC: National Academy Press. Nee, J., J. Ishimaru, and M. E. Hallenbeck. 2001. HOV Lane Performance Monitoring: 2000 Report. Seattle, WA: Washington State Transportation Center. Oregon Department of Transportation (ODOT). 1999. Oregon State Highway Transportation Volume Tables. Salem, OR: ODOT Transportation Systems Monitoring Unit.

Bertini, Leal and Lovell


ODOT Statewide Performance Measures

Table 2

Sample Corridor Performance Measures

Table 3

Sample Corridor Performance Measures


Site Map

Figure 2

Average Daily Traffic (ADT)

Figure 3

Average Speed

Figure 4

Travel Time

Figure 5

Percent Of Lane-Miles Congested

Figure 6

Lost Travel Time Due To Congestion

Figure 7

Demand Vs. Capacity

Figure 8

on-Recurring Congestion Example

Figure 9

Fusion Of AVL Data With Loop Detector Data


Bertini, Leal and Lovell


Potential Oregon Safety and Mobility Performance Measures

Safety 1. Maximize the safety of system operating conditions at all times a. Number of incidents on system per yearly VMT b. Customer perception of safety on transportation system 2. Minimize transportation conflict points for all modes a. Number of incidents/injuries near conflict points per number of conflict points b. Number of correctable crash sites funded for improvement 3. Improve the clarity and design of operations delivery to system customers. a. Annual survey questionnaire response b. Percent of system route miles of with basic/advanced/predictive travel time information available. c. Number of ATIS calls and website visits 4. Improve incident detection verification and response. a. Average time between notification and response/arrival and clearance b. Total duration of incidents

Mobility 1.

2. 3. 4.


Maximize system throughput a. PMT per system lane mile per station population b. VMT per passenger vehicle c. Urban/rural roadway miles at V/C greater the 0.70 by functional class d. Customer perception of quality of service, timeliness, and effectiveness e. Number or percent of signals verified or re-timed Improve travel time reliability a. Hours of delay for non-recurring congestion b. Delay per accident Minimize hours of delay experienced by customers a. Hours of stopped time per system user b. Hours of delay per system user Enhance inter-modal connections a. Number (or percent) of inter-modal connectors improved by operational strategies b. Accident rates at major inter-modal connections c. Time to access inter-modal facilities d. Customer perception of connection convenience Maximize modal choice a. Percent of state highways with bicycle lanes b. Percent of urban state highways with sidewalks c. Percent of total person miles of travel that are made in HOV lanes (on sections with HOV lanes) d. Percent of state residence with access to TDM program


Bertini, Leal and Lovell TABLE 2

Sta. Miles

1 2 3 4 5 6 7 8 9 10

1.22 1.63 1.07 0.7 1.4 1.33 0.96 1.18 1.03 0.47








105,800 192,900 177,900 100,000 290,300 325,700 273,500 242,500 253,000 127,200

157,000 286,000 263,900 148,300 429,600 481,900 404,700 358,800 384,400 191,900

2,088,900 3,106,600

Mobility Index

95 92 85 89 78 78 61 79 78 81





15,259,700 19,218,100 26,149,200 19,466,200 31,513,400 34,340,000 38,697,800 35,041,700 29,474,900 29,140,800

22,635,600 28,507,400 38,788,700 28,875,400 46,628,200 50,810,500 57,258,400 51,907,000 44,843,400 43,948,700


80,400 134,500 98,100 58,600 110,200 99,800 48,800 88,000 37,000 93,300

VMT by Congestion Level Percent Occupancy % 5-8 8-12 12-17 17-28 Nearcapacity Uncongested 11,900 7,300 800 0 28,400 19,000 3,300 100 29,100 27,600 8,200 3,400 17,600 15,400 3,700 1,100 46,200 60,700 29,200 23,500 54,000 74,100 28,700 29,900 23,200 66,000 88,700 38,600 35,800 47,500 26,700 16,100 17,600 42,700 84,500 60,500 23,700 7,600 500 9



Congested 0 0 0 0 4,000 1,400 1,400 800 10,500 1,600 30,900 2,200 3,200 900 13,700 3,300 8,800 1,800 0 0

PMT by Congestion Level Percent Occupancy % 0-17 17-28 >28

148,900 274,700 241,800 141,300 364,200 379,700 335,500 293,400 276,500 188,800

0 170 5,000 1,600 34,700 44,200 57,200 24,000 92,000 13

0 0 8,000 3,300 18,000 49,000 6,100 25,100 16,200 0

278,301,800 414,203,300 848,600 287,500 368,000 274,100 173,200 72,600 12,000 2,644,800 258,900 125,700

Bertini, Leal and Lovell






1 25,300 2 44,200 3 39,700 4 20,100 5 63,400 6 68,900 7 57,400 8 56,900 9 59,700 10 32,200 TOTAL 467,800 PERCENT

A 15 9 8 14 6 6 6 5 5 6 32,800 7%

Percent of VMT at LOS "X" B C D E 58 35 7 43 6 6 2 8 8 9 63,000 13%

26 56 64 44 9 4 10 9 4 15 91,300 20%

0 0 22 0 65 64 5 58 57 55 182,300 39%

0 0 0 0 14 20 65 20 25 15 91,000 19%

F 0 0 0 0 0 0 13 0 0 0 7,500 2%

>0.68 %

Total Delay veh-hr

Delay per VMT min/mi

0 0 22 0 79 84 83 78 83 70

8 2 502 247 745 580 184 557 209 135 3,170

0.02 0.00 0.76 0.74 0.70 0.51 0.19 0.59 0.21 0.25 4

Bertini, Leal and Lovell FIGURE 1


Site Map




Bertini, Leal and Lovell FIGURE 2


Average Daily Traffic (ADT)

160,000 MILEPOST Station 1 - 61.25 Station 2 - 62.47 Station 3 - 64.50 Station 4 - 64.60 Station 5 - 65.90 Station 6 - 67.40 Station 7 - 68.55 Station 8 - 69.31 Station 9 - 70.90 Station 10 - 71.37



ADT (vpd)







0 60







MILE POST Data Eastbound

ODOT two directions

Directional Distribution 50/50-ADT 1999


Bertini, Leal and Lovell FIGURE 3


Average Speed



2:55:20 pm 49 mph


14,000 58 mph


7:21:20 am

90 10,000

49 mph


36 mph


70 8:25:00 am Free flow speed


6,000 60 mph

59 mph


4,000 40 2,000

0 12:00 AM


3:00 AM

6:00 AM

9:00 AM

12:00 PM Time

Re-scaled cumulative Speed

3:00 PM

6:00 PM

9:00 PM

Average Speed

Best linear approximation on the curve where the slope is the speed


20 12:00 AM

Speed (mph)

V(x,t)- v0 t', vo=9500 mph per hour


4:01:20 pm

Bertini, Leal and Lovell FIGURE 4

Travel Time 70 Free flow travel time 7:36:20

Morning peak 50,000

11.67 min

Afternoon peak


Travel Time

25.92 min

Free-flow Travel Time

2:55:20 PM

Cumulative Travel Time



30,000 8:16:20 AM

11.68 min


7:21:20 AM 20,000 17 min


10,000 11.69 min 20


Free-flow Travel Time -10,000 12:00 AM

10 3:00 AM

6:00 AM

9:00 AM

12:00 PM Time

3:00 PM

6:00 PM

9:00 PM

Travel Time (minutes)



Bertini, Leal and Lovell FIGURE 5


Percent Of Mile Lanes Congested

100 90 80 70 60 50 40 30 20 10 0 12:00:00 AM

3:00:00 AM

6:00:00 AM

9:00:00 AM

12:00:00 PM


3:00:00 PM

6:00:00 PM

9:00:00 PM

12:00:00 AM

Bertini, Leal and Lovell



Lost Travel Time Due To Congestion







0 12:00:00 AM

3:00:00 AM

6:00:00 AM

9:00:00 AM

12:00:00 PM Time

3:00:00 PM

6:00:00 PM

9:00:00 PM

12:00:00 AM

Bertini, Leal and Lovell FIGURE 7


Demand Versus Capacity

4,500 1.00

4,000 3,500


3,000 2,500

0.60 V/C



1,500 1,000


500 0 TIME (hours) Station 1 Station 7

Station 2 Station 8

Station 3 Station 9

Station 4

Station 5

Station 6

Bertini, Leal and Lovell FIGURE 8


Non-Recurring Congestion Example

High congestion

Low Congestion

Low congestion

Bertini, Leal and Lovell FIGURE 9


Fusion Of AVL Data With Loop Detector Data

73 72 71

63.62 mi/hr

STATION 10 stop

14.61mi/hr 143





68 67


58.65 mi/hr

66 65

52.43 mi/hr STATION 5 STATION 4


63 62


Total travel time - Comet = 16.20 minutes Total travel time -Other vehicles = 11.29 minutes



Other Vehicles STATION 1

60 10:55:12 AM

10:57:12 AM


Slope of the curve which represents Average Speed

10:59:12 AM

11:01:12 AM

11:03:12 11:05:12 AM TIME AM

11:07:12 AM

11:09:12 AM

11:11:12 AM

Suggest Documents