Guidance on The Level of Effort Required to Conduct Traffic Analysis Using Microsimulation

CHAPTER 5.  BASE MODEL DEVELOPMENT

BASE MODEL DEVELOPMENT OVERVIEW

A systematic process is required to build a successful base traffic simulation model. The process should include clear documentation of the model structure and calibration criteria. Documenting and following the process ensures that the actual building of the model is done in an orderly way, thereby minimizing coding errors and mistakes.

Experience shows that speeding through the model building in the traffic software interface and relying strictly on the error-checking process to find all of the problems is not a cost effective approach. Although it takes time to set up and carry out the model, utilizing systematic procedures can make a big difference in the timeliness of completing the entire modeling analysis. The time dedicated to building the initial model is fairly small when compared to the overall modeling process. Taking the time to plan the modeling building and executing the work carefully will greatly reduce having to repeat work.

The base model development processes includes the following areas:

• Network (nodes, links, and link connectors).
  
  
• Lane geometry (number of lanes, length of turn lanes, etc.).
  
  
• Traffic control information.
  
  
• Travel demand:
  
  
  • Entering volumes.
    
    
  • Turning percentages.
    
    
  • O-D information.
    
    
• Error checking for all areas.

Procedures for developing base models have been documented in other guidebooks, such as the Traffic Analysis Toolbox Volume III: Guidelines for Applying Traffic Microsimulation Modeling Software and Traffic Analysis Toolbox Volume IV: Guidelines for Applying CORSIM Microsimulation Modeling Software.^(1,7) However, other guidebooks have not documented the development of O‑D matrices for use in simulation. The use of O-D matrices in traffic simulation models has grown and become increasingly important to produce effective and detailed analysis. The following section discusses the development of O‑D matrices.

There is a tradeoff in the level of effort required for preparing accurate O-D information and model quality and reliability. Generally, more complex networks and/or more complex geometric alternatives require more robust O-D efforts. These considerations must be taken into account while determining the scope of the project.

O-D MATRIX ESTIMATION

This section addresses the need to include O-D information into microsimulation tools and presents a number of methods for obtaining the O-D data.

Reasons to Use O-D Matrices

Historically, traffic simulation models were set up to have traffic demands entering from the exterior portions of the model study area and applying turning percentages at each junction. As a vehicle approached a junction, the stochastic assignment would use the turn percentages to assign a choice. This process would repeat until the vehicle left the network. The “look ahead” function (i.e., when a driver/vehicle becomes aware of an upcoming decision) varied by model. This function could have been defined by a particular distance or number of links. This approach is adequate in many but not all situations.

Figure 10 illustrates the path of vehicle A entering the system in the lower left corner and exiting the system in the upper right corner. At each junction, there is a turn percentage or volume that is used to determine the path the vehicle takes through the system. Note that in this figure, there are six decision points at which a vehicle gets assigned a path.

Traffic simulation models inherently cannot understand the conditions in the field. That is, vehicles do not know where to go unless the inputs are set up correctly in modeling terms. This is important, for instance, when a simulation model is intended to correctly evaluate weaving operations on a freeway.

In order for the model to correctly represent not only the traffic volume but also the patterns of movement, some type of O-D information is required. Using O-D matrices is particularly helpful in complex networks, especially when parallel facilities are included in the network. If the project purpose involves evaluating operational strategies and determining alternate paths based on congestion or other ITS-related information, then the use of O-D matrices makes the use of the model more efficient, especially when analyzing alternate geometric configurations in future networks. Once the O-Ds are established, it becomes easier to test design alternatives and other more complex strategies.

Every simulation software package has its own method for creating and inputting O-D information. Some software packages can do partial O‑Ds, O‑Ds on freeway facilities, or full zonal- or gate-based O‑Ds. Figure 11 and figure 12 show full O‑D-based models. The path of vehicle A is identified by the zone 3 to zone 10 O-D. (Note that the path of vehicle A is illustrated the same as vehicle A in figure 10.)

 
Figure 10. Illustration. Turn percentage O-D-based approach.

 
Figure 11. Illustration. Full O D-based approach.

 
Figure 12. Illustration. Alternate full O-D-based network configuration.

Advantages of Using Full O-D Matrices in Simulation

The process and procedures for developing a sound O-D matrix can be time consuming and complex. The collected field data, license plate surveys (if available), and a trip table available from the local MPO are used to develop an O‑D matrix for use in simulation. Approaches and techniques for developing O‑D matrices, known as O-D matrix estimation (ODME), are discussed in this section.

Once the O-Ds are established, the ability to test design alternatives and other more complex strategies becomes easier. For example, in figure 12, the previously illustrated freeway system is altered. The diamond interchange on the right is changed to a partial cloverleaf interchange. If the O‑D-based system of entering traffic demands is used, then the model only needs be altered geometrically and at the signal controls. The new assignment of trips automatically occurs, and the drivers/vehicles select the appropriate link leading to their destinations.

The advantages of using O-D matrices become even more evident in complex networks, especially when parallel facilities are included in the network. If the project purpose involves evaluating operational strategies and determining alternate paths based on congestion or other ITS-related information, then the full O-D matrix method is even more efficient. Figure 13 shows the same network (as shown in figure 11) with a parallel arterial north of the freeway. These types of complex models are not the focus of this report; however, this case illustrates additional advantages that can be gained in a microsimulation process when ODME techniques are used.

 
Figure 13. Illustration. O-D parallel facilities.

ODME Approaches

There are multiple methods for developing O-D inputs into microsimulation models. This is largely dependent on the software, which may have no O‑D inputs, partial O-D inputs, or full O-D inputs. Depending on the size, complexity, available data, and software platform selected for the project, the ODME technique may include one or a combination of the following techniques:

• Traffic counts.
  
  
• Surveys (license plate or roadside).
  
  
• TDMs.

ODME by Traffic Counts

For a small freeway segment or for basic freeway operations (e.g., ramp-to-ramp O-Ds), developing O-D matrices for simulation from traffic counts alone is a straightforward method that can be accomplished with a spreadsheet tool similar to what is typically used to reduce and balance traffic count data. Traffic counts at each on- and off-ramp are required, and mainline volumes between interchanges are required to account for possible errors in ramp counts or at locations where ramp counts are performed on different days.

The primary advantage of this methodology is that it is relatively quick and simple. One disadvantage comes to light when performing an analysis on future scenarios where the growth in traffic is determined in the form of growth percentage and applied to the existing counts. This method has become less desirable in the current state of the practice as it can either overestimate traffic in areas not expected to have growth trends in land use or it can underestimate growth in traffic by not taking into account regional growth and through traffic that may be attracted to the corridor when improvements are made.

Another disadvantage is that the zones for the O-D matrices are in the form of ramp termini or entry points, not true transportation analysis zones (TAZs). This means that for any future scenarios that have interchange or ramp reconfigurations, the O-D matrices will have to be modified by hand to reflect the new ramp termini or interchange reconfiguration. This introduces the potential for errors, and depending on the size of the system, it can be labor intensive.

ODME by Surveys

ODME for simulation can also be developed by conducting O-D surveys through license plate matching, driver intercept surveys, and newer technologies utilizing mobile source data. These procedures provide O-D data by matching a vehicle’s entry and exit points within the system (usually entry and exit ramps to a freeway). Since this methodology is a sample of the traffic within the system, a sampling plan must be prepared containing control counts at key locations within the study area to verify the level of traffic and to ensure that the vehicles included represent a statistically significant sample of the vehicles in the system. In addition to the survey, traffic counts are needed to develop expansion factors to expand the O-D survey data to match the traffic levels in the study area.

The advantage of using this method compared to traffic counts alone is that the O-D matrices for the analysis area are more accurate. The disadvantages are similar to those of the ODME by traffic counts mentioned in this section. An additional disadvantage is that this methodology typically includes only a sampling of the total traffic and may not provide information on through traffic on freeway segments.

ODME by TDMs

ODME for traffic analysis and microsimulation from existing TDMs has become popular as a result of the increased capabilities in software and computing power. However, O-D matrices cannot simply or directly be taken from a TDM and utilized in a microsimulation model. TDMs, however, can serve as the base platform for developing O-D matrices in terms of the network to be analyzed and TAZs to serve as O-D zones. This methodology offers several advantages, including the following:

• A larger study area can be considered beyond just the freeway segments and on- and off-ramps. This could include the local arterial system adjacent to the freeway and/or potential alternative routes to the freeway.
  
  
• The traffic analysis tool will be O-D based on true TAZs, not just entry and exit points to the freeway. This provides for a true dynamic assignment of traffic, which eliminates the need to manually reassign traffic for each network change/alternative as well as for future forecasts.
  
  
• Future forecasts can be derived at the TAZ level from accepted forecasts developed by the MPO and contained in the regional TDM at the TAZ level.

Disadvantages of this methodology include the following:

• ODME using TDMs requires an extensive data collection and calibration investment. Often, the TDM O-Ds are derived from regional surveys and estimates of local O-Ds for localized analysis of freeway sections. The process to derive the O-Ds in a TDM typically requires the same type of data that would be collected for the techniques of ODME by traffic counts and ODME by surveys. ODME can be calculated manually; however, it is typically more efficient to compute ODME using software. The level of effort and procedures can vary depending on the traffic analysis tool selected and the software platform of the regional model.
  
  
• Regional TDMs are validated to perform analyses of regional transportation plans and air quality conformity. This is usually done at the annual average daily traffic level (24-h assignment). This results in the need to process the ODME into at least hourly O‑D matrices, which may require a factoring process for peak spreading in congested regions for hourly, 30-min, and 15-min time intervals.

ODME Overview

There are several methods and approaches used to develop O-D matrices. Depending on network and study complexity, there are advantages and disadvantages to all of the methods, including the following:

• The direct measurement method may be simpler; however, data collection costs may be prohibitive and may limit the size of the data collection area. Additionally, the direct measurement method is not accurate for complex networks and for new geometric configurations.
  
  
• TDM-based techniques provide larger coverage and better traffic estimation in new geometric configurations; however, the accuracy may be reduced because of the coarse estimation techniques that are used to develop regional trip estimations.