Comparison of Simulation Models and The HCM: Title

Duplication for publication or sale is strictly prohibited without
prior written permission

of the Transportation Research Board.
Title: Comparison of Simulation Models and the HCM
Authors: Loren Bloomberg, CH2M HILL

Mike Swenson, The Transpo Group
Bruce Haldors, The Transpo Group
Word Count: 4343 words + 10 tables/figures (2500 words) = 6843 total words
Loren Bloomberg Mike Swenson Bruce Haldors

CH2M HILL The Transpo Group The Transpo Group
155 Grand Avenue, Suite 1000 11730 118th Ave NE, Suite 600 11730 118th Ave NE, Suite 600
Oakland, CA 94612 Kirkland, WA 98034-7120 Kirkland, WA 98034-7120
(510) 587-7720 (425) 821-3665 (425) 821-3665
(510) 622-9220 fax (425) 825-8434 fax (425) 825-8434 fax
lbloombe@ch2m.com MikeS@thetranspogroup.com BruceH@thetranspogroup.com
Submitted for consideration to:
Transportation Research Board

nd
82 Annual Meeting
January, 2003
Washington, DC
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
BLOOMBERG, SWENSON, HALDORS PAGE 2
ABSTRACT
Simulation modeling has become a standard approach for traffic operations analysis, supplementing (and in many
cases supplanting) the methodologies in the Highway Capacity Manual (HCM). The Concepts and Modeling
subcommittee of the Highway Capacity and Quality of Service (HCQS) committee has been undertaking a variety of
informal quantitative comparisons as part of its activities. This paper is an evaluation designed to increase
knowledge of the relationship between the HCM and simulation modeling.
The basic process for this effort was to develop an analysis test case with signalized intersections and a freeway
section that could be applied using the HCM and simulation, ask model developers to calibrate their simulation tools
to this test case, have them apply their models for future conditions, and then provide the results in a prescribed
method. This process was used as a means to replicate a typical modeling application, by a public agency or
consultant. A total of six (6) models were evaluated: CORSIM, INTEGRATION, MITSIMLab, Paramics, VISSIM,
and WATSIM.
The developers sent the results from the individual models to the authors, using a common spreadsheet format. The
authors compiled the results, completed some statistical analyses, and presented the results at the 2002 HCQS Mid-
Year meeting.
The results provided a rich dataset of comparative information for the simulation models and the HCM. From these
data and feedback from the participants, there were a number of conclusions from the process. In general, the
results indicated that the individual models tested produced reasonably similar results for the tested scenarios. There
was no indication that any one model was notably better or worse than the others, suggesting that model selection is
less important than the ability to effectively code, test, calibrate, and apply these models. There were differences
between the models and the HCM, but generally they were consistent with expectations and in the range of one LOS
grade (at most).
1. INTRODUCTION AND BACKGROUND

Simulation modeling has become a standard approach for traffic operations analysis, supplementing (and in many
cases supplanting) the methodologies in the Highway Capacity Manual (HCM) [1]. The Highway Capacity and
Quality of Service (HCQS) committee, charged with advancing the HCM methodologies, has recognized the
importance of simulation in Part V of the latest version of the manual.
Beyond Part V of the HCM, however, there is a recognized need for more information about the application of
simulation. Since most simulation models have some proprietary components, it is difficult to conduct a direct
comparison of various models on a theoretical level; application comparisons are needed. There are a number of
published papers that compare two or more models [2,3,4], and others that provide qualitative assessments of the
various models [5,6]. However, there are no comprehensive quantitative comparisons of a wide variety of models.
There are a number of challenges related to undertaking such a comparison. First, there are many simulation models
available (the SMARTEST project [7] counted 57 in a paper published in 2000), and many of the models change
frequently as developers improve their capabilities. Second, there are few (if any) individuals or groups who are
skilled in applying a large number of models. Finally, quantitative comparisons would be most appropriate if
undertaken by a sanctioned group, to ensure objectivity and independence. In committee meetings, the HCQS and
its subcommittees have considered various approaches for formalized model comparisons, perhaps even sanctioned
by the committee. All have been rejected, however, for a variety of reasons.
The Concepts and Modeling subcommittee of the HCQS has been undertaking a variety of informal quantitative
comparisons as part of its activities. The original assessments took place during the 2001 Mid-Year Meeting (in
Truckee, CA), but were not published. After that meeting, the authors were charged with updating the comparison
as part of activities planned for the 2002 Mid-Year Meeting (in Milwaukee, WI). This paper describes the process
and results of that effort.
The primary goal of the exercise was to increase knowledge of the relationship between the HCM and simulation
modeling. Some of the questions raised by the committee included the following:
• How do the results obtained using simulation modeling compare to an analysis using HCM methodologies?
• What is the range of results that can be expected for various simulation models?
• How do changes in demand affect the range and variability of simulation results?
• What techniques are used to calibrate the simulation models?
Also important were several goals that were not a part of the study. While some of these goals may be desirable,
they were explicitly excluded from consideration:
• Which simulation model is “best”?

• Which is better – simulation or HCM?
• Which model is easiest to calibrate?
• How should simulation be used “with” the HCM?
This last point is particularly important, because Part V of the HCM seeks to address that issue, to some degree.
However, the analysis described here is not prescriptive in any way; it merely compares the results of a set of
simulation models and the HCM.
In this paper, Section 2 describes the methodology employed in the study. Section 3 is a presentation of the results,
and Section 4 and Section 5 summarizes the observations and conclusions of the authors.
2. OVERVIEW OF METHODOLOGY
The basic process for this effort was to develop an analysis test case with signalized intersections and a freeway
section that could be applied using the HCM and simulation, ask model developers to calibrate their simulation tools
to this test case, have them apply their models for future conditions, and then provide the results in a prescribed
method.
The first step was to identify model developers to participate in the activity. A total of ten (10) model developers
were identified, based on commercial application of models (in the United States), applicability to the test problem,
and past participation in the HCQS committee. The developers of these ten models were contacted with a
description of the process, and asked for their willingness to participate. Six (6) of the model developers agreed to
participate. Table 1 summarizes the model developers, and lists the six models (CORSIM, INTEGRATION,
MITSIMLab, Paramics, VISSIM, and WATSIM) that were applied in the comparison. It is important to note that
none of the model applications were completed by authors; the modeling was completed by the developers (or their
designees) to ensure the highest level of knowledge and experience in model application.
The problem was developed to address common elements of traffic operations analysis, both for the HCM and
simulation. These elements include mainline freeway, ramps (merge, diverge, and weave sections), interchanges,
and surface streets (including signalized intersections). The problem set included a freeway with two interchanges,
and two cross-streets. The ramp terminal intersections were signalized, creating a set of four (4) intersections for
analysis in the network.
Previous experience with developing test cases suggested that the application problem should have the following
characteristics:
• Meaningful (representative of a typical traffic operations analysis in the field)

• Independent (a problem that could be analyzed using the HCM or any simulation model without prior
knowledge)
• Easy for developers (to ensure that the simulation analysis could be completed using available resources)
• Easy to analyze and compare (to allow fair comparisons to be made, using available resources)
• Relevant to HCM (comparison to the HCM were deemed to be an important part of the process and results)
The existing conditions data provided to the model developers included traffic volumes (balanced flows and turning
movements representing a typical day, but no origin-destination data), roadway geometry (generally sufficient for an
HCM or simulation analysis) and signal phasing/timing information. Also provided were performance data, using
HCM service measures: density on the freeway, and control delay at the intersections. Travel time and speed data
for existing conditions were also calculated. Note that the original problem was derived from a real-world dataset
(from a project in Idaho), but some of the input parameters were changed. Therefore, the performance data were
recalculated using the procedures in the 2000 HCM.
Model developers were instructed to calibrate their simulation to these existing conditions, using best practices.
Since there is no standard set of calibration guidelines (although FHWA is in the process of developing a set of
procedures for simulation), the developers used their own discretion in calibration. Most sought to minimize the
error between model output and the performance data (speed, density, delay, etc.).
Once they had calibrated their existing conditions models, they were asked to re-apply the calibrated models to
future conditions, using an updated set of traffic volumes. These demands varied by location, but were generally ten
to twenty percent higher than existing conditions. The model developers were not given future performance data,
and were asked to simply provide the results with these new volumes. The only other change in the future models
was that developers were asked to adjust the signal timing at the ramp terminal intersections, at their discretion.
The developers sent the results from the individual models to the authors, using a common spreadsheet format. The
authors compiled the results, completed some statistical analyses, and presented the results at the 2002 HCQS Mid-
Year meeting. In all of the presentations (including this paper), the individual model names were not provided. The
convention used was to name the six models as A, B, C, D, E, and F. These letters were assigned (using a random
order) to the six models applied.
This process was used as a means to replicate a typical modeling application, by a public agency or consultant. In
most cases, some (but not perfect) existing conditions data are available, and the individual applying the model is
able to calibrate to existing conditions. Also, while some estimate of future volumes is available, there is no
information about future performance (that is generally the reason for applying the simulation model). In other
words, the process applied here was not intended to represent how well simulation models would compare under
ideal conditions, but rather how well they compare when used in a typical application, where perfect information is
not available.
The specifics of the input problem are provided in the Figures 1, 2, and 3. As shown in Figure 1, the mainline
section consisted of three travel lanes in each direction. The arterial street system and ramp termini consisted of
two- lane ramps with four-lane arterials with turn pockets. Depending on the ramp terminus, some dual left-turn
lanes onto the freeway ramps were included. Under existing conditions, the mainline volume (Figure 2) varied from
4,500 vehicles per hour (vph) to 5,500 vph. In the future, the mainline volume increased to 4,900 vph to 6,200 vph.
An hourly profile was provided that outlined the peaking characteristics that should be assumed, based on a 0.95
peak hour factor (PHF). Although the developers were asked to collect data throughout the hour-long simulations,
the analysis of the results focused on the 30- to 45-minute window where the volumes were highest. This was done
to better match the HCM methodology as it relates to the PHF and an estimation of the peak 15-minute period.
As noted previously, this problem set originated from work conducted for a project in Idaho. In that particular
problem, the mainline segment only consisted of two-lanes in each direction. With the third lane added, additional
traffic was added to the system to achieve the near-capacity conditions desired for this data set. For the freeway, the
existing conditions were determined using HCM Chapter 22 methodologies, calculated using the individual chapter
methodologies for basic, merge, diverge, and weaving sections. Consistent with Chapter 22, Figure 3 reflects the
segment summaries of the various speed and density measurements for the existing conditions along the freeway
segments and ramp termini.
3. QUANTITATIVE RESULTS
3.1 Freeway Comparisons

Many different measures of effectiveness (MOE) can be collected for a system using simulation models. These
include, but are not limited to, vehicle queuing, vehicle and person delay, travel times, link evaluations (density or
headways), etc. This comparative analysis focused on three primary measures. These included travel time, average
speed, and lane densities. Since travel time is derived from the average speed, vehicle delay and lane densities, a
comparison between the travel times along various sections has not been included in this paper. Instead this paper
focuses on the average speeds and lane densities that are consistent in principal with the HCM calculations.
The freeway was divided into five primary sections. The sections were defined based on the methodology outlined
in Chapter 22 of the HCM. The diverge points and merge points are included in the sections, such that a basic
section could include the merge or diverge (or both) depending on the layout of the interchange. The sections are
defined in Figure 3. Given the amount of information collected in total, for the purposes of this paper the results for
just two key sections have been summarized. The two sections include results from the eastbound and westbound
direction of travel. These two sections were chosen since they highlight the variability in the models that occur as
the volumes increase in the future to a point that a roadway approaches or is at capacity.
The results for the six individual models for the two key sections are shown in Figure 4. For each section, the
density and average travel speed is highlighted. For each individual section or point, the results of the existing and
future operations are summarized. As shown in Figure 4, under existing conditions, the six models were able to
match reasonably well to the HCM results with respect to travel speeds and lane densities. For future conditions
with increased volumes, the six models showed an increase in the variability and standard deviation. This variability
is most observed in the western section (eastbound direction). This section has a combination of high mainline
volumes and a significant diverge volume as well. This combination creates oversaturated conditions in the future
that impact the system operations. This comparison shows that as the traffic demands get closer to the capacity
there is more variability in the model results. Although the models do not report LOS directly, the results from the
model have been correlated to the HCM for purposes of this comparison. Given the importance of LOS for
decision-making (especially when predictions are LOS E or LOS F), it is important to understand whether or not the
variability in these model results equate to a different LOS for that section.
For most of the comparisons in Figure 4, the simulation models predicted a LOS with the same grade or one
different from the HCM. For the congested section on the west end of the network, there are some data points that
are two LOS grades different from the HCM.
3.2 Intersection Comparisons
In addition to the travel speeds and lane densities summarized for the mainline, information was collected for the
operations of the ramp termini. A summary of the intersection delay as reported by the six-models compared to the
HCM results are shown in Figure 5. Under both existing and future conditions, there is large variability in the
reporting of an intersection average delay. Based on discussions with model developers and other attendees at the
HCQS mid-year meeting, there are two potential causes:
• Future signal timing was not provided, developers were asked to optimize the timing based on the
simulation.
• The HCM definition of intersection control delay may be different than the definition and methodology
used by some of the models. This requires additional review of the individual models.
Similar to the freeway comparisons, local jurisdictions tend to use LOS as a guide for measuring the impact of
improvements or determining if a development project will be approved. As shown in Figure 5, depending on the
location and horizon year, the models were generally within one LOS grade of the HCM, with a maximum delay
difference of about 15 seconds. A broad summary is that the difference between the HCM and the simulation
models averaged one-half of a LOS grade.
3.3 Summary
Although the goal of the study was not to say whether the HCM or simulation is more accurate, it was a goal to
understand how the data collected through the simulation models compared to the calculated results using the HCM
methodologies. To illustrate these differences, the individual model results were all compared to the HCM
calculations. Table 2 highlights the percent difference between the results extrapolated from the simulation models
and the HCM value; with the values reported as absolute. A review of this information indicates that generally for
the freeway sections, the largest differences between HCM and the simulation models occurred in the eastbound
direction for the section that was operating at or above capacity (as defined by the HCM).
A comparison of the intersection delays puts into perspective the results displayed in Figure 5. The differences
between the simulation models and HCM range from the similar results to differences as high as 89 percent.
4. OBSERVATIONS
In addition to the comparisons between individual models, some higher level analysis of the results was undertaken.
These were completed to better understand the general comparison between simulation and the HCM.
4.1 System Level

The first comparison was at the system level, using system measures of vehicle-miles traveled (VMT) and vehicle-
hours traveled (VHT). These measures are frequently used by decision-makers, but are difficult to calculate using
HCM methodologies. Table 3 summarizes the results for all six models using these measures.
For VMT, the models (on average) predicted an increase in VMT of about 11 percent (from 57,000 to 63,000) from
the existing to the future scenario. This was not a surprising result, because the traffic volumes were precisely
specified. What was more surprising was the variability of the results – the standard deviation in VMT was about
8,000 vehicle-miles between models. For example, the range for existing conditions was 45,000 to 65,000 vehicle-
miles. The low variability in the ratio of future/existing VMT indicates that the models are internally consistent
(i.e., a model that predicts a low existing VMT will also predict a low future VMT). However, it does cast doubt on
the reported absolute values of VMT.
Similar results were found using the VHT measure. The models were consistent in predicting that total travel time
will increase faster than the travel distance (the expected result), but the variation in VHT predictions was relatively
high, even though the models were internally consistent.
4.2 Freeway
Table 4 presents a comparison of the travel time estimates from the six models. The data in the table are the ratio of
travel time – future vs. existing. For example, if the westbound travel time prediction was five minutes for existing
conditions and six minutes for the future scenario, a ratio of 1.2 was reported. The Table includes data for all six
models. Travel time was selected as a typical high-level measure used by decision-makers.
For the westbound direction, which has minimal congestion, all six models predicted moderate increases in freeway
travel time for the future scenario. Five of the six models predicted travel time increases of less than ten percent; the
sixth was only thirteen percent.
For the eastbound direction, however, the results were much more varied. Three of the models predicted only
moderate increases in travel time (less than ten percent), while the other three suggested large increases in travel
time (34 to 96 percent). These results are indicative of the level of congestion in the eastbound direction, and the
challenges in estimating performance when demands are greater than capacity. While all of the models are designed
to handle oversaturated conditions, a particular challenge with calibrating simulation models is estimating capacity,
because capacity is not an input to most models. With slight changes in input parameters, the resulting capacity can
change significantly, and the performance data will vary as well. This is clearly illustrated in the variability of the
data in Table 2.
4.3 Intersections
Figure 6 illustrates the variability of the results for signalized intersection analysis. The Figure includes data for all
six models for the four intersections evaluated in the study. The individual data points represent the predictions for
existing conditions delay (on the horizontal axis) and future delay (on the vertical axis). With the increased volumes
in the future scenario, the expectation is that the best fit line (displayed on the graph) would have a slope greater
than one (i.e., future delay is higher than existing conditions delay). The graph indicates this result did occur,
although the correlation is relatively low (r-squared value of 0.56). The variation in the data points is higher for the
intersections with more delay, suggesting the predictions have more variability when there is more congestion,
consistent with the freeway findings.
These data are more challenging to interpret, however, for several reasons:
• There was variation in the signal timing plans, a major driver for predicting delay, as the developers created
their own future signal timing plans.
• The freeway congestion in the future scenario may meter traffic to the intersections, reducing delay for the
future scenario (this occurs for several data points).
• The definition of control delay is interpreted differently for every model, and is likely less consistent across
models as delays and queues increase.
The points above do not invalidate the results, because the point of the exercise was to replicate typical “real-world”
modeling conditions. It is clear from the findings that care must be taken in interpreting the output from individual
simulation models for predictions of delay at individual intersections.
5. CONCLUSIONS
5.1 Implications
This study was not intended to be the definitive comparison of simulation models and the HCM; this is probably an
impossible goal. Rather, the goal was to advance the process of comparing models and provide more information to
the HCQS committee. Based on feedback from the committee and others, that goal was successfully achieved.
The authors’ interpretations of the results suggest the following implications:
• A clear definition of “calibration” and calibration data is enormously important in the use of simulation models.
For instance, a more precise definition of where data were collected and what elements have the most priority to
calibrate to could potentially bring the different simulation model results closer together.
• For this study, the results indicated that the individual models tested produced reasonably similar results for the
tested scenarios. There was no indication that any one model was notably better or worse than the others. This
suggests that model selection is less important than the ability to effectively code, test, calibrate, and apply these
models.
• There were differences between the models and the HCM, but generally in the range of one LOS grade (at
most). While this does not prove that models can be successfully applied for problems well beyond HCM
methodologies, it does provide a measure of comparison between simulation and the HCM.
• The ability of simulation models to make absolute predictions for system measures (e.g., VMT), travel time, or
delay is uncertain. While the models were generally consistent in their relative predictions (i.e., future vs.
existing), there was a notably high degree of variability in their absolute predictions (i.e., model A vs. model B)
for some measures. This is also to say as the level of detail is increased, the certainty with the results is not
directly proportional and may in fact be inversely proportional at times.
5.2 Limitations of the Study

It is important to recognize the limitations of this study, as follows:
• Number of data points: The performance data only included point data on each segment, and complete detector
information (e.g., speed, occupancy, and/or volume) was not available.
• Definitions: This was most significant for control delay, which is defined differently by each developer (i.e.,
how delay is measured upstream of a signal).
• Accuracy of existing conditions data: Since these data were fictional (although adapted from a real-world
study), there was some question about their accuracy. However, since they were checked by hand and some
values were calculated, they may have been more accurate than some real-world datasets.
• Level of congestion: There are likely larger differences between the models and the HCM at higher levels of
congestion; the example in this study only had moderate levels of congestion (demand/capacity ratios of 1.00 to
1.10).
• Calibrating below capacity: Some developers noted that it was difficult to calibrate the level of congestion
without being able to determine the capacity. To precisely determine capacity, existing conditions with
oversaturated segments would be needed.
• Not enough situations to analyze: It was suggested that additional cases (e.g., creating a “build” future
scenario) would have provided more comparisons between the models.
While all of these points are valid, it should also be noted that they are the norm in typical applications of
simulation.
5.3 Feedback from the HCQS Committee

These findings were presented to members of the HCQS committee and others, at the 2002 Mid-Year meeting. The
format included a formal presentation of the results by the authors, and then an open discussion forum. The open
discussion included several invited panel members (a mix of public agency technical staff, consultants, and model
developers) as well as some HCQS committee members and interested others. The feedback from that group
included the following:
• Most were interested in the results, and felt the work was worth doing.
• Many HCQS committee members commented that the “real-world” approach was best, in that differences in
model results might be attributed to imperfect information. One comment was that these models were applied
by the “world experts” (i.e., the model developers), so that more variability should be expected with the range
of skills of the general community.
• There were some comments identifying specific improvements that could be made to the study design.
• There was mixed opinion about whether these changes were important (i.e., to create a more controlled study)
or non-critical (i.e., because the study was representative of real-world applications).
• There was general (although not universal) agreement that the results should be made available to a larger
group. Those in agreement felt that this was an important topic on which to disseminate information. The
minority felt that the study design limitations suggested caution in publishing findings.
5.4 Recommended Next Steps

At this point, the next step in this research is not clear. Discussion of these results will continue at the 2003 TRB
Annual Meeting, and additional investigations may be suggested. It is hoped that dissemination of these findings to
the user community will result in guidance for additional research. Based on the HCQS feedback, some possible
next steps include the following:
• An investigation of the reasons for differences between the individual models (e.g., analysis of car-following
models)
• Additional case studies, with real-world or other examples
• Investigation of the results with changes to the supply or traffic volumes
• An updated analysis with better calibration data (more data points, origin-destination data, and more congestion
for assessing capacity)
REFERENCES
[1] Transportation Research Board. Highway Capacity Manual. Special Edition, 1997.
[2] Wang, Y. and P. D. Prevedouros. Comparison of CORSIM, INTEGRATION and WATSim in Replicating
Volumes and Speeds on Three Small Networks. Transportation Research Record 1644, pp. 80-92, National
Research Council, Washington, D.C., 1998.
[3] Bloomberg, L. and J. Dale. A Comparison of the VISSIM and CORSIM Traffic Simulation Models on a
Congested Network. Transportation Research Record 1727, pp. 52-60, National Research Council, Washington,
D.C., 1999.
[4] Prevedrouros, Panos D. and Y. Wang. Simulation of a Large Freeway/Arterial Network with CORSIM,
INTEGRATION, and WATSim. Transportation Research Board 78th Annual Meeting, January 1999.
[5] Skabardonis, Alexander. Assessment of Traffic Simulation Models : Final Report. Berkeley, CA :
University of California, Institute of Transportation Studies, 1999.
[6] Elefteriadou, L., J. Leonard, H. Lieu, G. List. Beyond the Highway Capacity Manual: A Framework for
Selecting Simulation Models in Traffic Operational Analyses. Transportation Research Board 78th Annual Meeting,
January 1999.
[7] ITS, University of Leeds. “SMARTEST Final Report for Publication”,

http://www.its.leeds.ac.uk/projects/smartest/finrep.PDF, January 2000.
TABLE 1
Invited Participants
Simulation Model Developer Contact

CORSIM* FHWA Gene McHale
INTEGRATION* Virginia Tech Hesham Rakha
MITSIMLab* MIT Tomer Toledo
Paramics* Quadstone Scott Aitken
VISSIM* ITC Thomas Bauer
WATSIM* KLD Associates Mark Yedlin
AIMSUM TSS N/A
FREQ UC Berkeley N/A
Synchro Trafficware N/A
Transmodeler Caliper Corporation N/A
*Participating Model
TABLE 2
Model/HCM Comparison Summary
EXISTING (% Difference) FUTURE (% Difference)

Model A B C D E F A B C D E F
Freeway Speeds
EB west of Wisconsin Off 11 4 4 8 2 1 14 6 42 42 3 66
EB Wisconsin Off to On 1 3 8 0 2 4 18 28 66 50 2 69
EB Wisconsin on to Milwaukee Off 12 9 9 14 3 10 25 10 24 12 13 2
EB Milwaukee off to on 2 2 3 3 5 2 1 4 1 1 6 3
EB east of Milwaukee On 9 7 4 3 1 3 13 8 1 3 4 4
WB east of Milwaukee Off 5 2 3 3 2 1 8 2 4 18 3 9

WB Milwaukee off to on 4 1 4 1 3 1 1 2 2 57 0 6
WB Milwaukee on to Wisconsin Off 13 5 6 6 1 8 16 6 12 18 7 11
WB Wisconsin Off to On 4 1 3 2 5 2 0 1 1 1 1 1
WB west of Wisconsin On 10 4 5 4 1 2 14 5 4 1 2 8
EXISTING (% Difference) FUTURE (% Difference)

Model A B C D E F A B C D E F
Intersection Delay
Milwaukee WB Ramp Terminal 13 50 29 21 35 13 19 59 33 7 39 15
Milwaukee EB Ramp Terminal 7 14 38 0 44 10 14 48 55 38 56 36
Wisconsin WB Ramp Terminal 0 5 14 32 10 27 25 38 34 19 2 3
Wisconsin EB Ramp Terminal 1 36 31 29 42 3 82 84 85 74 89 79
TABLE 3
Systemwide Summary Comparison
Individual Model Average Standard

Measure Scenario A B C D E Deviation
Vehicle- Existing 44,500 57,800 57,200 60.900 65,300 57,000 7800
Miles Future 50.300 66,000 62,900 64,900 72,700 63,000 8200
Traveled Future/Existing Ratio 1.13 1.14 1.10 1.07 1.11 1.11 0.03
Vehicle- Existing 890 1200 1190 1330 1270 1180 170
Hours Future 1080 1420 1530 1830 1500 1470 270
Traveled Future/Existing Ratio 1.21 1.19 1.28 1.37 1.18 1.25 0.08
TABLE 4
Freeway Summary Comparison
Ratio of Future/Existing Travel Time

Simulation Model Eastbound Westbound
A 1.06 1.05
B 1.07 1.02
C 1.48 1.02
D 1.34 1.13
E 1.07 1.03
F 1.96 1.07
FIGURE 1
Sample Problem Geometric Data
FIGURE 2
Sample Problem Traffic Volume Data
FIGURE 3
Existing Conditions MOE – Calibration Points
FIGURE 4
Freeway Model/HCM Comparison
FIGURE 5
Ramp Termini Delay Comparison
FIGURE 6
Intersection Summary Comparison
Intersection Delay (Existing vs. Future)
40
35
Future Delay (sec)
30
25
20 y = 1.0679x
R2 = 0.5636
15
10
10 15 20 25 30 35
Existing Delay (sec)

Comparison of Simulation Models and The HCM: Title

Uploaded by

Copyright:

Available Formats

Comparison of Simulation Models and The HCM: Title

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Comparison of Simulation Models and The HCM: Title

Uploaded by

Copyright:

Available Formats

Duplication for publication or sale is strictly prohibited without

prior written permission

Title: Comparison of Simulation Models and the HCM

Authors: Loren Bloomberg, CH2M HILL

Loren Bloomberg Mike Swenson Bruce Haldors

Submitted for consideration to:

Transportation Research Board

1. INTRODUCTION AND BACKGROUND

• Which simulation model is “best”?

• Meaningful (representative of a typical traffic operations analysis in the field)

3.1 Freeway Comparisons

4.1 System Level

5.2 Limitations of the Study

5.3 Feedback from the HCQS Committee

5.4 Recommended Next Steps

[7] ITS, University of Leeds. “SMARTEST Final Report for Publication”,

Simulation Model Developer Contact

EXISTING (% Difference) FUTURE (% Difference)

WB east of Milwaukee Off 5 2 3 3 2 1 8 2 4 18 3 9

EXISTING (% Difference) FUTURE (% Difference)

Individual Model Average Standard

Ratio of Future/Existing Travel Time

Intersection Delay (Existing vs. Future)

You might also like