VISSIM Modeling Guidance 9-12-2017
VISSIM Modeling Guidance 9-12-2017
VISSIM Modeling Guidance 9-12-2017
This document outlines MDOT SHA best practice methodologies for VISSIM related
microsimulation operational analyses and recommended modeling techniques.
This document is not a tutorial for VISSIM; rather it provides guidance on specific concerns previously
noted by the TFAD staff. Engineers applying these methodologies should already be familiar with the
latest VISSIM software package, currently at version 9.0 (as of 9/12/2017).
Users should also be familiar with the latest Highway Capacity Manual (HCM), Maryland Manual on
Uniform Traffic Control Devices (MD MUTCD), and general transportation vernacular to ensure
accurate engineering judgment during the modeling process.
A minimum of one week must be taken into consideration for TFAD staff to review and confirm the
VISSIM models are acceptable at each stage of the modeling effort (i.e. a minimum of three review
periods with additional review periods if more than one Build condition is modeled).
Additional time may be required and must be discussed with the TFAD staff to ensure project schedule
adherence. TFAD staff will review the models for accuracy per the VISSIM checklist.
Modeling Techniques
Vehicle Inputs
Vehicle inputs must reflect current vehicular composition and speeds using existing vehicle traffic
counts and travel data. At a minimum, Vehicle Inputs will take into consideration automobiles and
trucks as two separate vehicle classes (exception: routes with no truck access). Additional breakdown of
vehicle classes are appropriate if data is available (i.e. motorcycles, medium trucks vs. heavy trucks etc).
For all project studies multiple Vehicle Input types must be created for all roadways entering the project
area. For example, side streets with no trucks might use 100% automobiles, whereas mainline streets
might use 90% automobiles and 10% trucks.
If the project is a transit oriented study, bus volumes should not be included in the Vehicle Input. Bus
“volumes” will be input as Public Transit (PT) frequencies. If the study allows for a mix of known and
unknown transit, the modeler can consider the unknown bus volume as a Vehicle Input and the known
bus “volume” as a PT frequency.
Segments of roadway with turning bays should be modeled as links with all lanes accessible, rather than
multiple parallel links (Scenario 1) each associated to a turning movement, as shown below, unless the
existing conditions include a physical barrier between turn lanes. TFAD recognizes this approach differs
from the PTV modeling technique. However, this approach allows users to then model forced lane use
(with the use of no lane change options) through connectors if necessary (Scenario 3) or allow vehicles
to merge smoothly into the turning bay (Scenario 2). Generally, this approach works best for longer
turning bays, but for consistency, all models should use the “one link-all lanes” approach and adapt as
needed.
Merges and diverges with acceleration and deceleration lanes should be modeled similarly (one link-all
lanes), one link with the acceleration or deceleration lane included as part of the mainline link, as shown
below, unless the existing conditions include a physical barrier between the mainline and the ramp lanes
(ex. Collector-Distributor lane).
Width exaggerated
In general, parallel link modeling is not an accepted methodology for TFAD operational analysis using
VISSIM software unless specific roadway geometry prohibits movement along the lane (ex. solid
barriers), ramp design allows for single on/off access from the freeway (ex. tapered diverge/merge), or
the modeler can provide field data to show that all drivers merge/diverge using the taper only. There
may be case by case exceptions, but the modeler should consider the above one link-all lanes approach
unless the conditions suggest otherwise.
All connectors should be short and should not significantly overlap over the two links it connects.
The purpose of this section is to provide VISSIM modelers with a set of baseline driving behaviors, which
can be applied to different link segments, in order to more effectively help with the calibration process of
VISSIM models for the Maryland Department of Transportation State Highway Administration (MDOT
SHA). The driving behaviors defined below were compiled, evaluated, and summarized from a pool of
most frequently used VISSIM driving behavior models obtained from industry experts who frequently
perform VISSIM analyses for MDOT SHA.
The goal of this compilation is to assist VISSIM modelers. All base behaviors can still be modified to
better reflect the specific corridors being analyzed; however, this list could aid in beginning the calibration
process and the Travel Forecasting and Analysis Division (TFAD) recommends utilizing the suggested
ranges of parameters defined within the document.
PTV VISSIM uses a psycho-physical car following model which is stochastic and discrete developed by
Wiedemann nameed Wiedemann 74 and Wiedemann 99, further discussed below.
• CC0 Standstill distance: Desired distance between the rear-bumper to front bumper of the
stopped cars. This parameter has greater impact to maximum flow rate when the traffic is in jam
conditions.
• CC1 Headway time: The distance in seconds that the following driver desires to maintain with
the lead vehicle. Note: this parameter is defined as a time distribution starting from VISSIM 9.0.
Model versions prior to 9.0 will result in conversions of the time distribution to a static value that
may impact your overall model.
Note that desired safety distance=CC0+(CC1* speed)
• CC2 (Following variation): How much more distance than the desired safety distance
(CC0+CC1) before the lagging driver intentionally moves closer to the lead vehicle.
Wiedemann 74 model
The Wiedemann 74 car following model is most frequently applied to arterial segments. It has three
parameters which can be modified to simulate real world traffic conditions:
• Average standstill distance: Average desired distance between two stopped cars.
• Additive part of safety distance: this value affects the computation of desired safety distance.
• Multiplicative part of safety distance: both the additive value and multiplicative value are used
for computation of the desired safety distance. The greater this value, the greater the distribution
of safety distances increases.
Note: For the vehicle to identify the downstream routing decision points, the option “Combine Static
Routing Decision” located under “Static Vehicles Routes” must also be selected.
Cooperative Lane Change
It is recommended this option be selected for all behaviors, as it smooths transitions into more realistic
driving behaviors.
Additional Factors
Additional factors that influence driver behaviors include the look back distance at connectors, the
roadway grade combined with truck vehicle compositions (if used), and general vehicle fleet with their
associated desired speeds. These factors are all considerations for additional “tweaking” to enhance
calibration accuracy during VISSIM modeling.
Suggested Range
Default
Parameter Unit Weave/Merge/Diverge
Value Basic Segment
Segment
CC0 Standstill distance 4.92 feet 4.5-5.5 >4.92
CC1 Headway time 0.9 sec 0.85 to 1.05 0.8 to 1.5
CC2 ‘Following’ variation 13.12 feet 6.56 to 22.97 13.12 to 39.27
Threshold for entering
CC3 -8 Use Default
‘following’
Negative ‘following’
CC4 -0.35 Use Default
threshold
Positive ‘following’
CC5 0.35 Use Default
threshold
Speed dependency of
CC6 11.44 Use Default
oscillation
Oscillation
CC7 0.82 feet/sec2 Use Default
acceleration
CC8 Standstill acceleration 11.48 feet/sec2 Use Default
Acceleration at
CC9 4.92 feet/sec2 Use Default
50mph
Reminder: the goal of this compilation is to aid VISSIM modelers in the calibration process. The
parameters of the driving behaviors listed in Table 4 below can be changed to better reflect specific
corridors. It is recommended, however, that the driving behavior values changed fall within the suggested
ranges previously discussed in Tables 1 through 3. Driving behaviors should be visually verified within
the simulation to ensure the behavior represents realistic operations.
Table 4: Driver Behavior Model Names
Freeway
101 Freeway Basic Conservative I
102 Freeway Basic Conservative II
103 Freeway Merge Conservative
104 Freeway Weave Conservative
105 Freeway Aggressive I
ARTERIAL
Description Name # LINK TYPE # Name Description
Model can be used for simulating Aggressive arterial segments.
Model is used for simulating conservative driving on arterial segments. Arterial Basic
Significant factors include SDRF at 0.10, car following model
Throughput reduction observed was around 15%. The lane change Conservative I 201 205 Arterial Aggressive I
parameter values are low and maximum cooperative braking value is
parameters values are kept low and SDRF is 0.60
BASIC also high
Throughput reduction is higher than the above model (around 18%). Higher throughput is achieved when compared to the above model.
Arterial Basic
Maximum deceleration is kept high and look ahead distance value is 202 206 Arterial Aggressive II Model simulates aggressive behavior with values of lane change
Conservative II
high parameters being high.
Model can be used for conservative driving on arterial segments where
reduced throughput is desired. Throughput reduction observed was Merge Arterial
203 MERGE -
around 13% Additive and multiplicative part of safety distance values Conservative
Note: Arterial Aggressive I & II above are also suitable for arterial merge/weave segments to
are 2.30 and 3.4 respectively
increase throughputs and simulate aggressive lane changing.
Model can be used for conservative driving on weaving arterial
Weave Arterial
segments. Model provides low throughput, around 18% with high 204 WEAVE -
Conservative
weaving.
All percent changes by behavior types are estimates and may vary depending on the scenario they are applied to.
Transit
For all Transit, bus alighting and boarding should be considered in addition to bus travel times,
schedules, capacity (vehicle types), all stop locations, etc. TFAD currently models an on street bus stop
as 50 feet to 100 feet depending on urban density. An alternative to boarding/alighting data is to use
dwell time information, though this must be supported with field verified information.
Speeds
Turning Speeds
Turning speeds for intersection movements, or tight left/right turning vehicles, should be modeled using
the speed distributions provided below. These speeds differ from the Synchro defaults and are based on
MDOT SHA data.
Speed reduction zones should be placed at the sharpest point on the curve of the link or connector. The
speed reduction zones for turning movements should be short, usually within 5-15 feet depending on the
curve length. Excessively long reduced speed zones will reduce the turning movement volume capacity
and should only be used if the turning movement excessively reduces vehicle throughput.
Wide left turning movements or free right movements where vehicles can travel faster are especially
susceptible to this condition and can be modeled with higher turning speed distributions with longer
speed reduction zones (e.g. 5-30 feet at 25mph), if appropriate.
Speed reduction zones for ramps, specifically loop ramps, should use a distribution of the ramp caution
speed limit, usually within the 30-45 mph range. These can span the entire ramp (ex. tight loops) or only
the sharpest curve of the ramp (ex. slip ramps) depending on field data.
Mainline Speeds
Mainline desired speeds should be modeled as a distribution of existing speeds along the corridor, not as
the posted speed limit. Vehicles modeled in VISSIM must reflect existing conditions as accurately as
possible. Scenario analysis may be performed after the base calibration in complete; however, existing
conditions must be reflected in the models.
When first opening a VISSIM model, care should be taken when converting from KPH to MPH (i.e.
when converting from metric to imperial). Do not switch to imperial units and keep the speed as-is; this
will result in unrealistic speed distributions.
Currently, TFAD uses the default VISSIM Maximum Acceleration and Deceleration distributions. Make
note should these be altered in the modeling effort.
Note that conflict zones work most efficiently for non-congested locations and tight conflict areas. For
wide turns, congested networks, and other complex facilities, priority rules may be more appropriate to
allow for smoother traffic flow.
Signals
Signal timings should use RBC NEMA phasing standards or VAP for complex/innovative signals. All
signal timings must use MDOT SHA, County, or City timing sheets. New signals must meet MDOT
SHA standard practice and the RBC timing sheet must be supplied to TFAD for review.
Permissive left turn signal heads can be coded as an “Overlap” with parent phases as the through and left
movement combined due to vehicles in Maryland operating as though left turn yellows are permissive.
This movement can also be coded through the “Or signal group” option in the Signal Head tool, but
vehicle throughput should be evaluated to make sure the capacity of the turn is maintained (i.e. vehicles
in Maryland tend to turn on yellow arrow, which operates more like the overlap condition than the “Or
signal” condition).
Right Turn On Red (RTOR) conditions must be coded into the networks where vehicles are permitted to
turn if the signal is red. To code RTOR, use the stop sign tool and under the “RTOR” tab, select the
“Only on Red” option for the appropriate Signal Controller and Signal Group. The stop sign should be
positioned on the link/connector performing the right turn while a signal head for the through movement
should still be coded in on the through link.
Caution: Import of Synchro files into VISSIM can lead to multiple errors and should be done with
caution. Always confirm Synchro timings with actual controller timing sheets when possible – TFAD
staff is trained in reviewing signal controller timing sheets and will request corrections to signal timings
if they do not match the controller outputs.
12 Maryland Department of Transportation State Highway Administration
Updated: August 2017
Routing
Static Routing
TFAD currently uses Static Routing for most VISSIM simulation models. This requires a balanced
network of traffic volumes to input in the VISSIM model that must be approved by TFAD. Routes
should start at the farthest point from a “split” or volume change location to ensure the most distance for
vehicles to make a decision.
Caution is advised for interchange locations where routing might cause “loop” conditions where a
vehicle will be removed from the highway only to return in the opposing direction. To avoid these
conditions, push highway traffic at interchanges through the following intersections rather than stopping
a route right after the end of the ramp movement, as shown below.
Route end points must be on the same link as the following route’s start point.
Breakdown of truck routes versus automobile routes, or route combinations will be left at the discretion
of the modeler. However, methodologies are expected to be submitted to TFAD for review.
Dynamic Routing
Dynamic routing should be discussed on a case by case basis with the TFAD staff as this requires an
additional macroscopic modeling effort.
Two calibration metrics are required of all VISSIM models submitted to TFAD:
- Travel time and/or speed
- Vehicle throughput
Additionally, engineering judgment will be required for locations with existing queues and overall
network operations. All calibration must consider the following:
- Seeding time must allow a car to travel from one end of the network to the other; customary
simulation seeding times span from 900 seconds (15 minutes) to 1,800 seconds (30 minutes).
Longer seeding times should be considered for excessively large networks or high congestion.
- A minimum of 5 simulation runs must be completed before average outputs of all runs can be
used for analysis. Additional runs may be necessary, up to 15 runs or by showing convergence of
the model.
Calibration of the network using travel times or speed must report short segment data in addition to
overall corridor travel time/speed. TFAD requires a ± 10 percent travel time variation for small
segments (no more than 1 mile long) and ± 5 percent travel time variation over the entire corridor
analyzed. Exceptions permitted on a case by case basis with justification.
For a facility spanning more than 1 mile, it is recommended to break the facility into segments based on
obvious breakpoints (ex. between signalized intersections, or at ramps). These new smaller segments
would then be calibrated at ± 10 percent variation with an overall corridor calibration of ± 5 percent.
To calibrate to travel times or speeds, floating car runs or collected speed data may be used (ex.
RITIS.org probe data). This may result in two separate data sets: one from floating car runs, and one
from an outside source. Do not mix the calibration of travel times from floating car runs with speeds
collected from an outside source. Two options are available if multiple data sets are available:
1. Average the speed data with the travel time runs into one data set (i.e. convert speeds into travel
time runs or vice versa and calibrate the VISSIM outputs to the average of the two),
2. Use only one data set, either travel time runs from the floating car runs, or the speed data from an
outside source, and keep the other data source for validation.
The volume calibrations should not exceed 10% of the count traffic volume and/or GEH<5.
Caution: A frequent error noted is the use of the balanced traffic volume network for calibration of a
VISSIM model. This is an incorrect calibration method. Calibration should not be made using the
demand volume (i.e. the balanced volume network), rather they should meet the throughput measured in
the field (i.e. raw data count).
Calibration sheets are required for review and must be presented with the start of stage (2) of the
VISSIM schedule (See Quality Review and Schedule). Example calibration tables are provided below.
Transit oriented studies must include transit travel times separate from automobile travel times (ex.
bus/tram/light-rail).
Queues
Maximum and average queue lengths should be collected at the stop bar locations for signalized
intersections or stop signs. Queues occurring on freeways should be measured from the start of the
queue by observing the simulation and determining the start point. Networks should be modeled such
that the maximum queue length measurements are encompassed by the network and queues do not
extend past the end of the link.
Measuring delay per vehicle should consider the HCM categories for LOS grade. Node delays at
unsignalized versus signalized intersections are different and must be evaluated correctly.
Node “start of delay segment” should consider the length of the queues at that node. Alternatively, this
may be zeroed out if the edges of nodes from nearby intersections are bordering the node (i.e. back to
back node systems).
Once data is collected from the VISSIM model, total intersection delays should be translated from the
latest Highway Capacity Manual (HCM) to a letter grade LOS.
For all studies the link(s) on which the merge, diverge, or weave occurs must be evaluated for density
output. Note that the modeler should use HGV and auto densities to translate to a LOS, and must not
solely translate the VISSIM “All Vehicle” density to a LOS grade. TFAD allows for the use of a 2.5
factor to convert HGV density to passenger car per mile per lane (pcpmpl), which is added to the auto
density, and then converted to a LOS grade per the HCM breakdowns. Similar to intersection LOS
categories, freeway segments have different breakdowns for weaves, diverges and merges, which must
be considered when reporting LOS.
Delays at the diverge/merge/weave may be considered in addition to the density; however, a delay
estimation using node or travel times may not be translated to a LOS using HCM delay tables.
Benefit cost analysis (BCA) may be considered for VISSIM models that consider Build scenarios. Each
Build scenario would include a change in the network, resulting in changes to vehicular delay.
TFAD’s current approach is to determine the delay variation at the location of the Build change using
either travel time segments or nodes. Due to the software limitations, some node systems may be too
large to encompass an entire interchange, for example, and thus travel times may be used.
TFAD does not currently use network wide delay as a means for BCA due to scale of projects; however,
this MOE may be considered for very small networks (independent intersections, single interchange etc)
and is needed for the Network Performance Measures, below.
Deliverables
The required deliverables of a VISSIM modeling effort to MDOT SHA’s TFAD include:
- All VISSIM models and associated VISSIM files (ex. RBC and VAP files), for each stage of the
schedule (see Quality Review and Schedule)
- Calibration tables (see Calibration)
- A Calibration and Methodologies Memorandum including version of VISSIM used and detailed
volume diagrams used in the network,
- MOEs of final (i.e. not the base model) VISSIM networks (see Measures of Effectiveness)
- Full technical memorandum with all results.
-end-
For questions, comments, or suggestions regarding this report please contact Ms. Carole Delion via
email at cdelion@sha.state.md.us or Ms. Lisa Shemer, Division Chief, Travel Forecasting and Analysis
Division.