Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

CORSIM, PARAMICS, and VISSIM: What the Manuals Never Told You

Fred Choa, P.E., Ronald T. Milam, AICP, David Stanek, P.E.

Abstract. With the increasing use of microsimulation traffic software in operations


analysis, the need to identify which tool to use and the ability of the software to provide
traditional traffic engineering measures of effectiveness consistent with the 2000 Highway
Capacity Manual (2000HCM)(1) has become a major area of debate. This paper will
provide a comparison of the three major traffic simulation software programs in use today
and the results of the evaluation matrix developed by the authors.

CORSIM, developed by the Federal Highway Administration (FHWA), is one of the most
commonly used micro-simulation programs for modeling vehicle traffic operations. In
addition, it is fairly common practice to create CORSIM models using Synchro, a traffic
signal optimization and coordination software developed by Trafficware.

PARAMICS, developed by Quadstone Limited, a Scottish company, is a software used to


model the movement and behavior of individual vehicles and transit on local arterial and
regional freeway networks. But unlike the other microscopic simulation programs,
PARAMICS is UNIX-based. But through the use of the Microsoft Windows-based software
Exceed by Hummingbird, the X-Windows platform can be emulated on a standard
workstation computer.

VISSIM was developed by Planung Transport Verkehr (PTV), a German company.


Innovative Transportation Concepts (ITC) based in Corvallis, Oregon distributes and
supports VISSIM throughout Northern America. VISSIM is one of the latest
microsimulation software programs available and provides significant enhancements in
terms of driver behavior, multi-modal transit operations, interface with planning / forecasting
models, and 3-D simulation.

This paper describes the following key factors that may affect the decision on which
software to use for a specific project involving a freeway interchange and the consistency of
the model output to traditional traffic engineering measures of effectiveness.

• Model Development (i.e. input requirements and coding effort);


• Calibration to Field Conditions (i.e. driver behavior, traffic flow characteristics, and
traffic control operations);
• Validation Requirements (i.e. travel times, queue lengths and level of service);
• Animation (i.e. graphics, viewing options, and backgrounds); and
• Model Output and Consistency with the 2000 HCM.

I. INTRODUCTION

The purpose of this paper is to provide transportation planners and traffic engineers with a
comparison of the set-up requirements, model development effort, and analysis results for
a typical freeway interchange study using the CORSIM, PARAMICS, and VISSIM
microsimulation traffic software programs. This paper describes the following key factors
that may affect the decision on which software to use for this type of project and the
consistency of the model output to traditional traffic engineering measures of effectiveness.

• Model Development (i.e. input requirements and coding effort);


• Calibration to Field Conditions (i.e. driver behavior, traffic flow characteristics, and
traffic control operations);
• Validation Requirements (i.e. travel times, queue lengths and level of service);
• Simulation / Animation (i.e. graphics, viewing options, and backgrounds); and
• Model Output and Consistency with the 2000 Highway Capacity Manual (HCM).

II. DESCRIPTION OF TRAFFIC SIMULATION PROGRAMS

Traffic simulation models are used in many cases to visually display analysis results and
provide a system-wide analysis instead of isolated components (i.e. intersections, ramp
merge / diverge, weaving sections, etc). Other common reasons for their use are to
analyze complex or unique roadway geometries that other analysis programs cannot
evaluate or to evaluate a combined system of arterial and freeway facilities. Three
commonly-used traffic simulation programs in the United States are CORSIM, PARAMICS,
and VISSIM.

CORSIM, developed by the Federal Highway Administration (FHWA), is one of the most
commonly used simulation programs for modeling vehicle traffic operations. In addition, it
is fairly standard practice to create CORSIM models using Synchro, a traffic signal
optimization and coordination software developed by Trafficware. CORSIM is one of the
programs included in the TSIS 5.0 software package. The package also includes TRAFED,
a network editor, and TRAFVU, an animation software.

The TRAFVU program provides a two-dimensional simulation using rectangular shapes for
cars, trucks, and buses. The colors for vehicles and the background can be adjusted using
a limited color palette. Pedestrians are not shown in the animation, but do effect the
operation of turning vehicles (although vehicles look like they are waiting for no reason).
Various performance measures can be shown as part of the animation.

PARAMICS, developed by Quadstone Limited, a Scottish company, is a program used to


model the movement and behavior of individual vehicles and transit on local arterial and
regional freeway networks. But unlike the other microscopic simulation programs,
PARAMICS is UNIX-based. But through the use of a Microsoft Windows-based software
called Exceed by Hummingbird, the X-Windows platform is emulated on a standard
workstation computer.

The PARAMICS Modeler program combines the network editing and visual animation tasks
and allows for simultaneous editing and simulating. Both two-dimensional and three-
dimensional animation can be performed using enhanced rectangular shapes for cars,
trucks, buses, and trains. The model elements can be given a wide variety of colors, and
the vehicle shapes can be customized. Additionally, a graphic file (such as an aerial
photograph) can be shown as the background for the animation. A separate program,
PARAMICS Analyzer, is used to visually display performance measures.

VISSIM was developed by Planung Transport Verkehr (PTV), a German company.


Innovative Transportation Concepts (ITC) based in Corvallis, Oregon distributes and
supports VISSIM throughout Northern America. VISSIM is one of the latest
microsimulation software programs available and provides significant enhancements in
terms of driver behavior, multi-modal transit operations, interface with planning / forecasting
models, and 3-D simulation.

VISSIM provides animation capabilities similar to PARAMICS, with major enhancements in


the 3-D simulation of vehicle types (i.e. from different passenger cars, trucks, transit
vehicles, light rail and heavy rail). In addition, movie clips can be recorded within the
program, with the ability to dynamically change views and perspectives. Other visual
elements, such as trees, building, transit amenities and traffic signs, can be inserted into
the 3-D animation.

The software acceptance issue is important to note because each simulation model is built
upon a set of stochastic algorithms that attempt to represent traffic flow through various
types of network systems under various conditions. This paper does not evaluate the
different algorithms but does point out that reviewing agencies such as the California
Department of Transportation (Caltrans) have been more likely to accept the use of
simulation models that were built upon traffic flow conditions measured and modeled in the
United States. But similar to the entire microscopic traffic simulation field, opinions are
changing fast as products like VISSIM and Paramics are used more and more here in the
US.

III. DESCRIPTION OF STUDY AREA

The study area for this evaluation is shown in Figure 1 on the following page. Depicted in
the figure are the geometrics, traffic control, and future (2025) peak hour traffic volumes
associated with a proposed improvement project for the U.S. Highway 50 / Missouri Flat
Road interchange near the historic gold mining town of Placerville, California. The existing
interchange operates poorly and suffers from limited spacing between signalized
intersections. After considering a variety of interchange improvement alternatives, a single
point urban interchange (SPUI) design was selected as the recommended future
interchange configuration. The SPUI configuration is unique in California, with only a
handful of SPUI interchanges in operation. Therefore, extensive traffic operations and
design analysis was required before local and State agencies accepted the design as the
preferred project design in this location. The use of traffic simulation analysis tools was
instrumental in the traffic operations analysis and served as the basis for this paper.
As shown in the figure, the study area extends east to include the U.S. 50 / Placerville
Drive / Forni Road interchange. This interchange was included so that realistic weaving
operations could be analyzed between the two interchanges, which are approximately ½
mile (800 meters) apart and connected by a bridge across Weber Creek canyon. The
vertical sag on the bridge compounds the problem of the short weaving section length
between the interchange by introducing grades of up to plus or minus eight percent for the
interchange ramps. Although the original analysis included a substantial number of
analysis locations, this paper focuses on a comparison of simulation model results for the
following locations:

• U.S. 50 / Missouri Flat Road ramp terminal intersections;


• U.S. 50 / Missouri Flat Road eastbound off-ramp;
• U.S. 50 / Missouri Flat Road westbound on-ramp; and
• U.S. 50 westbound and eastbound weaving sections between the Missouri Flat
Road and Placerville Drive / Forni Road interchanges.

IV. TRAFFIC SIMULATION MODEL SET-UP

As part of the traffic simulation model comparison, initial set-up times for the U.S. 50 /
Missouri Flat Road interchange project were documented. Set-up times were compared
because they are a good indicator of user-friendliness, learning curve requirements, and
structure of the graphical user interface (GUI). In general, the set up times were similar
with 3 to 4 days required to get each model coded and debugged. PARAMICS and VISSIM
have more input parameters, which can improve the accuracy of the models, when
compared to CORSIM.

For example, the path-based routing in VISSIM was used to better reflect lane utilization
through the SPUI. The PARAMICS network editing tools were used to match the proposed
turn radii at all intersections and ramp junctions. This type of refinement was not possible
in CORSIM due to the link-based trip assignment and 50-foot minimum spacing
requirement between nodes. As a result, an additional day was used to refine and
complete the final set up for the PARAMICS and VISSIM models of the interchange.

V. TRAFFIC SIMULATION MODEL CALIBRATION AND VALIDATION

Calibration is the process by which the individual components of the simulation model are
refined and adjusted so that the simulation model accurately represents field- measured
and observed traffic conditions. The major components or parameters of a simulation
model that require calibration include the following:

• Traffic control operations;


• Traffic flow characteristics; and
• Driver behavior.
This step should not be confused with validation; whereby the accuracy of the model is
tested by comparing traffic flow results generated by the model with field collected data.
Validation is discussed later in the paper and is directly related to the calibration process
because adjustments in calibration are often necessary to improve the model’s ability to
replicate field measured traffic flow conditions (i.e. validation).

With regards to calibration, traffic simulation models contain numerous variables to define
and replicate traffic control operations, traffic flow characteristics, and driver behavior.
Simulation models contain default values for each variable, but also allow a range of user-
applied values for each variable. In some cases, the variables affect the entire network
while others are specific to individual roadway segments or nodes. Changes to these
variables during calibration should be based on field-measured or observed conditions. In
other words, a change in the variables should be justified and defensible.

Unfortunately, the user manuals for simulation models provide little or no information about
the source or appropriateness of the default parameters, nor do they provide substantial
guidance on how the user should modify these parameters for different types of conditions.
Therefore, the user has a greater responsibility for ensuring that appropriate changes are
made that are based on field-measured data and not exclusively on engineering judgment.

Under ideal conditions, the calibration of individual components of a simulation model will
improve the model’s ability to replicate traffic flow results that match field conditions within
an acceptable range of error. Typical traffic flow characteristics that can be used in
validation include traffic volumes, average travel time, average travel speed, queue lengths,
and density. Unfortunately, professional guidelines that define the acceptable range of error
for these characteristics have not been developed. Instead, transportation professionals
have either ignored the need for validation or developed their own guidelines. Examples of
validation guidelines used in recent projects by the authors and accepted by agencies such
as Caltrans are contained in Table 1. Although these guidelines are a starting point for
discussion, they lack statistical justification to determine if they provide an acceptable range
of error.

VI. ANALYSIS PROCEDURES

Because of each software’s unique capabilities, default input parameters, and the
stochastic nature of simulation programs, the same input information regarding roadway
geometrics, intersection configurations, traffic control, and traffic volumes does not yield the
same results. Calibration can help to minimize the potential differences between programs.
However, many projects involve scenarios where transportation facilities are proposed in
areas where they do not exist today or where transportation conditions are projected to
change so dramatically that existing conditions are not expected to represent future
conditions. Additionally, the method of calculating the performance measures differs
between software programs and none of the programs provide a level of service analysis
that is directly comparable to the 2000 HCM methodologies.

Table1
Validation Guidelines

Parameters Description Validation Criteria

Percent difference between input volume 95 to 105 % of


Volume Served
and the simulation model output or assigned volume observed value

Standard Deviation between floating car average travel


Average Travel 1 Standard
times and simulated average travel time for a
Time Deviation
series of links

Standard Deviation between floating car average travel


Average Travel 1 Standard
speed and simulated average travel speed for individual
Speed Deviation
links

Percent difference between observed freeway density (from 90 to 110 % of


Freeway Density volume counts and floating car travel speed) and simulated observed value
density

Average and
Percent difference between observed queue lengths and 80 to 120 % of
Maximum Vehicle
simulated queue lengths observed value
Queue Length

According to the 2000 HCM methodology, the level of service at an intersection is based
upon average control delay on the approaches. The output files for CORSIM provide the
control delay for each network link and the total delay for each turn movement. PARAMICS
only provides total delay, not control delay, for each link and not for each turn movement.
VISSIM is the most flexible and provides two ways to measure delay. The total delay can
be measured between any two points in the network, or can be measured for each
intersection turn movement using the node evaluation process.

For freeway analysis, the 2000 HCM methodology uses freeway segment vehicle density to
determine level of service. All three software programs allow for the measurement of
density at a point on a link or for an entire link. CORSIM does not provide a direct estimate
of density for each lane although the other two programs do. The density by lane is
needed in ramp junction analysis to determine the density in the ramp influence area on a
freeway with three or more lanes.

The U.S. 50 / Missouri Flat Road interchange improvement project described above is a
good example of these problems and presented the authors the opportunity (and
challenge) to conduct this simulation model comparison. In general, each program was
used to set up a simulation model of the study area shown in Figure 1. The models used
for comparison were limited to p.m. peak-hour vehicle traffic operations under 2025
conditions since higher traffic volumes were projected for the evening peak hour. In order
to determine the effects of random seed numbers for the stochastic models, five runs were
conducted with each simulation model and the averages for performance measures were
summarized and compared. The results of the analysis are discussed in the next section.

V. ANALYSIS RESULTS SUMMARY AND COMPARISON

As stated above, the analysis results included the following locations:

• U.S. 50/Missouri Flat Road ramp terminal intersections;


• U.S. 50/Missouri Flat Road eastbound off-ramp;
• U.S. 50/Missouri Flat Road westbound on-ramp; and
• U.S. 50 westbound and eastbound weaving sections between Missouri Flat Road
and Placerville Drive/Forni Road.

Performance measures were extracted from each simulation model to determine the
corresponding level of service (LOS) based on 2000 HCM procedures and methodology.
Table 2 below compares the ramp terminal intersection LOS results while Table 3
compares the ramp and weaving section LOS results. Although the LOS results are
similar, key differences are worth noting.

For example, the PARAMICS model had lower values for percent demand served, which
was caused by congestion through the SPUI associated with incorrect lane utilization. The
problem stems from northbound vehicles on Missouri Flat Road destined for the westbound
on-ramp being in the incorrect lane prior to reaching the SPUI intersection. This type of
problem occurs because CORSIM and PARAMICS use link-based routing and is solved in
VISSIM due to the use of path-based routing. PARAMICS currently has a limited look-
ahead distance which allows vehicles to anticipate turn movements up to two links away.
According to PARAMICS technical support, upcoming upgrades to PARAMICS will
eliminate this type of problem. A similar problem also occurs in CORSIM but can be
mitigated somewhat through the use of conditional turn movements. But this function is
only limited to adjacent links.

In addition, sensitivity tests were conducted to better understand how minor changes in
demand volume would affect the operations of the SPUI and the results of the microscopic
traffic simulation. The simulation results for the ramp junctions were compared with
analysis results from the Highway Capacity Software (HCS) to determine whether the
results are comparable. The traditional HCM methodology has greater acceptance by
many traffic engineers and transportation planners as long as the traffic volume inputs fall
within acceptable ranges. Table 4 contains the HCS analysis results.
Table 2
2025 Intersection Operations Summary – 2025 Conditions

Percent
Demand Demand Delay
Intersection Volume Served (sec./veh.) LOS
CORSIM
Missouri Flat Road/Prospector’s Plaza Drive 3,333 98% 21 C

Missouri Flat Road/U.S. 50 Westbound Ramps 3,056 98% 5 A

Missouri Flat Road/SPUI 3,978 98% 28 C

Missouri Flat Road/U.S. 50 Eastbound Ramps 4006 96% 3 A

Missouri Flat Road/Mother Lode Drive 4128 99% 14 B

PARAMICS

Missouri Flat Road/Prospector’s Plaza Drive 3,333 95% 22 C

Missouri Flat Road/U.S. 50 Westbound Ramps 3,056 95% 4 A

Missouri Flat Road/SPUI 3,978 93% 36 D

Missouri Flat Road/U.S. 50 Eastbound Ramps 4006 93% 6 A

Missouri Flat Road/Mother Lode Drive 4128 86% 26 C

VISSIM

Missouri Flat Road/Prospector’s Plaza Drive 3,333 98% 20 C

Missouri Flat Road/U.S. 50 Westbound Ramps 3,056 98% 5 A

Missouri Flat Road/SPUI 3,978 100% 30 C

Missouri Flat Road/U.S. 50 Eastbound Ramps 4006 100% 3 A

Missouri Flat Road/Mother Lode Drive 4128 100% 9 A

After reviewing the HCS results and the graphic simulations for each simulation model, the
following conclusions were reached:

• Weaving results from CORSIM underestimated density and LOS. For example,
even using the latest version CORSIM (i.e. TSIS 5.0) the CORSIM model interface
node (between NETSIM and FRESIM networks) at the westbound on-ramp from
Placerville Drive / Forni Road creates an artificial bottleneck that artificially meters
traffic entering the freeway which creates gaps in the platoons from the on-ramps
that do not actually occur. This situation creates unrealistic traffic flows through the
westbound weaving section.
Table 3
2025 Ramp Junction and Weaving Operations Summary – P.M. Peak Hour
(Traffic Simulation Results)
Density
Ramp Junction/Weaving Section (1) LOS
CORSIM
U.S. 50/Missouri Flat Road eastbound off-ramp 17.3 B
U.S. 50/Missouri Flat Road westbound on-ramp 69.4 F
U.S. 50 westbound weaving section - Placerville Drive/Forni Road
18.2 B
to Missouri Flat Road
U.S. 50 eastbound weaving section - Missouri Flat Road to
21.4 C
Placerville Drive/Forni Road
PARAMICS
U.S. 50/Missouri Flat eastbound off-ramp 39.4 E
U.S. 50/Missouri Flat westbound on-ramp 102.6 F
U.S. 50 westbound weaving section - Placerville Drive/Forni Road
33.4 D
to Missouri Flat Road
U.S. 50 eastbound weaving section – Missouri Flat Road to
36.6 E
Placerville Drive/Forni Road

VISSIM

U.S. 50/Missouri Flat eastbound off-ramp 26.4 D

U.S. 50/Missouri Flat westbound on-ramp 59.0 F


U.S. 50 westbound weaving section - Placerville Drive/Forni Road
30.4 D
to Missouri Flat Road
U.S. 50 eastbound weaving section – Missouri Flat Road to
26.6 D
Placerville Drive/Forni Road
Notes:
(1) Density is defined as passenger cars per mile.

Table 4
2025 Ramp Junction Operations Summary – P.M. Peak Hour
(HCS Results)

Ramp Junction/Weaving Section Density (1) LOS


U.S. 50/Missouri Flat eastbound off-ramp 36.0 E

U.S. 50/Missouri Flat westbound on-ramp 33.0 D

Notes:
(1) Density is defined as passenger cars per mile.
• The PARAMICS and VISSIM models did not have this problem and provided more
reliable simulations of freeway on-ramps, ramp junction and weaving sections.

• Ramp junction results from CORSIM underestimated the density and LOS for the
eastbound off-ramp. The 2025 freeway volume approaching this ramp is 3,165
vehicles in two lanes of which 1,155 vehicles are diverging to the off-ramp. LOS B
as calculated based on the CORSIM results is not consistent with these volumes.

VI. CONCLUSIONS

The traffic operations results for this specific project should not directly applied to all
interchange projects because of the unique design of a SPUI and the wide variety of input
parameters and their ranges. However, this comparison paper does point out the
importance in choosing the right simulation analysis tool and determining the applicability of
simulation model results when using only one model.

The following conclusions summarize the results of comparing three of the most used traffic
simulation programs.

• CORSIM provided the shortest set-up time. Both PARAMICS and VISSIM required
about an additional day for model refinement.

• Both CORSIM and PARAMICS use link-based routing which can result in inaccurate
lane utilization for closely-spaced intersections. The path-based routing in VISSIM
eliminates this problem.

• Both PARAMICS and VISSIM provide three-dimensional animation although VISSIM


has more options for enhancing the visual setting. The CORSIM two-dimensional
animation is more simplistic.

• No software provides average control delay for each turn movement although
CORSIM does provide average control delay for each approach. All models report
total delay by link.

• CORSIM has an artificial barrier between arterial and freeway networks that can
cause inaccuracies such as the “metering” of traffic on high-volume on-ramps or
“backups” of traffic on high-volume off-ramps.

• Overall, PARAMICS and VISSIM generated simulation results that better matched
field observed conditions, traffic engineering principles, and expectation / perception
of reviewing agencies including Caltrans.
The results show substantial differences between the three microscopic traffic simulation
models. If only one model were applied, the user would likely accept the results unless the
graphic simulation displayed unexpected results. Reviewing the results from all three
models raises the obvious question, “Which model is correct?” Or, the question could be
phrased as, “Are any of the models correct?”.

The user manuals do not include information about the thoroughness of software testing for
various types of projects or situations. In other words, software documentation or user
manuals do not provide the information traffic engineers and planners need to answer the
critical questions or to determine the reliability of the simulation results.

Fred Choa, P.E. (ITE Member)


Ronald T. Milam, AICP (ITE Member)
David Stanek, P.E. (ITE Associate)
Fehr & Peers Associates, Inc.
2990 Lava Ridge Court Suite 200
Roseville, CA 95661
(916) 773-1900 phone
(916) 773-2015 fax
www.fehrandpeers.com

ENDNOTES

(1) Transportation Research Board, Highway Capacity Manual (HCM), Fourth Edition,
2000.

You might also like