13.5.3 Calibration Parameters and Best Practices

Calibration demands several iterations of adjustments to calibrate to real-world conditions. There are several model parameters that are typically attuned to the collected field data. The calibration of a model relies on the data collected. presents various calibration measures and potential data sources
Table 13-7: Various Calibration Measures and Potential Data Sources
Calibration Measure
Facility
Potential Data Sources
Potential Output Files
Volume/Throughput
Freeway/Ramps
Tube or video counts
Link evaluation, Data collection points
Arterials/Intersections
Tube, video, or manual counts
Link/node evaluation, Data collection points
Travel Times
Freeways/Arterials
Field Travel Time Runs or Probe Data
Travel time segments
Speed/Congestion
Freeways/Arterials
Spot-Speed Data Collection or Probe Data
Data collection points, Travel time segments
Bottleneck Locations
Freeways/Arterials
Field photographs/videos/notes or recent peak period aerial imagery (if available)
Visual inspection of model
Microsimulation calibration best practices are summarized below:
  • Select key performance measures for each project. Calibrate selectively for those performance measures.
  • Use reliable observed data for the performance measures used for calibration.
  • Calibrate based on study area dynamics (bottleneck throughput or duration) and time-variant (travel time, speed) performance measures.
  • Use a representative day for calibration rather than an average day that combines multiple days.
The following sections outline possible parameters for calibrating Vissim, CORSIM, and Trafficware’s SimTraffic models.

13.5.3.1 Vissim Calibration

Several parameters can be adjusted in Vissim to fully calibrate a model. The most impactful and common parameters are outlined below:

13.5.3.1.1 Lane Change Parameters

Two common parameters related to lane change behavior are the LCD and the Emergency Stop Distance (ESD). LCD is the point at which a vehicle attempts to change lanes prior to a decision point (e.g., a turning movement). The default LCD setting of 656.2 feet may cause vehicles to switch lanes too late. ESD is defined as the location where vehicles decide to stop and wait for a lane change. These two parameters have great effects on upstream traffic, so their calibration greatly improves a model’s accuracy. Additionally, the following parameters are calibrated in Vissim; reference the Vissim user guide for more guidance:
  • Advanced Merge;
  • Safety Distance Reduction Factor;
  • Cooperative Lane Change; and
  • Maximum Deceleration of Own Vehicle (MDOV)

13.5.3.1.2 Speed Distributions

The speed distribution curve in Vissim is set to represent the speed distributions from the collected field data. The posted speed limits are defined as 85 percent of the desired speed distribution in Vissim. Maximum speed distributions is recommended to be set no greater than 10 mph above the speed limit and no less than 5 mph below the speed limit. When calibrating speed distributions, it is possible to increase the percentage of vehicles traveling at the speed or the maximum speed limit cap based on field data.

13.5.3.1.3 Driving Behavior

The interaction between vehicles is modeled using the Wiedemann 1974 and 1999 carfollowing models. Such models are assigned for different driving behavior containers. The Wiedemann 1974 is used for ‘Urban (motorized)’ link types that represent urban arterial roads and streets. The Wiedemann 1999 is primarily used for ‘Freeway (free lane selection)’ link types to model freeway operations. With updated parameters, Wiedemann 1999 is often used for ‘Cycle-Track (free overtaking)’ links to model bicycle lanes. The link type ‘Footpath (no interaction)’ does not monitor the interaction between users of this link type. As such, it can be used for pedestrian movements and walkways and crosswalks. shows all the carfollowing parameters in a Vissim model that can be changed. However, it is recommended to only modify CC0, CC1, and CC2 when adjusting the car-following parameters. Car following and lane changing parameters are shown in .
Table 13-8: Car-Following and Lane Changing Parameters
Parameter
Description
CC0
Standstill distance. The distance between two vehicles when they are not moving
CC1
Following distance. A time distribution of the speed-dependent portion of safety distance.
CC2
Longitudinal oscillation. Defines the distance in which a driver will intentionally move closer to the car it is following.
CC3
Perception threshold for following. Defines when the beginning of the deceleration process occurs.
CC4
Negative speed difference. Low values result in more sensitive reactions when following vehicles.
CC5
Positive speed difference. Low values result in more sensitive reactions when following vehicles.
CC6
Influence speed on oscillation. Defines how car-following distance impacts the acceleration of vehicles.
CC7
Oscillation acceleration value. Minimum value used when a driver is following another vehicle.
CC8
Acceleration value when starting from a standstill.
CC9
Acceleration value at 80 km/h
Diffusion Time
This is the maximum amount of time that a vehicle can wait at an ESD before the vehicle is removed from the network
Min. Clearance
The minimum distance between two vehicles after a lane change occurs
Safety distance reduction factor
The safety distance of a following vehicle after a lane change, the safety distance of the vehicle making the lane change, and the distance to the preceding, slower lane change.
Maximum deceleration for cooperative braking
Defines how much a trailing vehicle will brake to allow an adjacent vehicle to change into its lane.
Use implicit stochastics
When this option is checked, safety distance, desire acceleration, desired deceleration, minimum lateral distance is stochastic to reflect variations in human perception. When this option is not checked, the characteristics listed above are non-variable.