This section presents more detailed information on freight models and their potential various features. Just as a database may have many different tables that contribute to the whole, a freight model is often composed of many different features that, when taken together, cannot be easily categorized, although straightforward categorizations such as the one here aid comprehension.
The discussion presented here was developed by considering the structure of various freight models that the authors have developed, implemented, or otherwise encountered in practice. We examined many real-world freight models that agencies or others use, studying the theoretical and empirical underpinnings of each. The resulting typology summarizes the findings of this exercise, focusing especially on the way that different model components can be developed. The focus here is on standalone models, as opposed to more elaborate integrated systems (e.g., long distance commodity-based freight model combined with a tour-based commercial vehicle model; our typology treats these as two separate model types).
Freight Models: Use Cases
When designing, implementing, or applying freight models, modelers should be aware of who will use the model output, what questions will they ask of the model (such as these examples), and what data inputs, processes, and granularity or resolution will be needed to address the question. Availability of resources for model development, maintenance, and application are additional considerations, but are not discussed in depth here.
User types include cities, counties, regions, states, corridor coalitions, and nations. Worldwide organizations may also use freight models, for example to study the impacts of ocean shipping.
When users ask questions of the model, they normally expect a response that can be characterized by:
- The form of the response needed, or specific metrics typically required (list)
- The resolution needed:
- Spatial
- Temporal
- Industry
- Commodity type
- Modes
- Vehicle size/weight, engine type
- Trip purpose/context (urban)
- Cargo or payload attributes
Freight Models: Model Features
Theory: The fundamental assumptions about the structures, behavior and dynamics of the transport system are provided by theories from various relevant disciplines like economics (e.g., input/output structures; game theory); management (e.g., supply chain organization); physics (e.g., maximum entropy, gravity); and psychology (e.g., decision-making, psychometrics). A special class of models are the data-driven models: these make no a priori assumptions about the system but instead will replicate its observed behavior. Machine Learning models are an example of a data-driven approach. Important concepts throughout different theories relate to the representation of time and behavior through time, uncertainty, system equilibrium, and network formation.
Methods: Quantitative approaches use methods from economics, statistics, optimization, and other fields. Models use analytical methods with equations to calculate outcomes. Econometric methods allow analysts to estimate the value of unknown model parameters that represent observed behavior. If complexity in the system (i.e., heterogeneity, uncertainty or dynamics) is relevant, one can use advanced methods or discrete simulation. Simulation models can be event-driven or time-driven, with or without explicit agents. System dynamics and agent-based models are examples. Optimization methods can be important to model networks or demand-supply equilibrium. When describing behavior, however, we should also account for the selection of suboptimal choices.
Data: The empirical foundation consists of the data used to develop model parameters and validate structures (i.e., ensure that the model performs reasonably well from a prediction standpoint). Input data sources are a form of measurement of a given element. Therefore, the list of elements that are being measured is an important characteristic of the empirical foundation. Example data elements include employment by industry, commodities by type, survey responses, and truck counts.
Granularity or resolution is an important attribute of both input and output data:
- Scale – macro, meso, micro (agent)
- Spatial
- Temporal
- Industry
- Commodity type
- Modes
- Vehicle size, engine types
- Trip purpose/context
- Cargo or payload attributes
For output data, it is also important to establish what metrics can be reliably derived from model results.
Minimum Model Features for an Example Use Case
The following table illustrates how model features can be specified to meet the needs of a specific use case, focusing on resolution needed and mechanisms to consider. The use case in question is a truck parking needs assessment by a multi-state corridor coalition. The proposed theoretical basis is a Shipper-Receiver-Carrier since the coalition is most interested in truck trips between origins (shippers) and destinations (receivers). The coalition is interested in the following impacts:
- Flows (volumes by weight, value, number of moving units; resolution: area, O/D, network)
- Asset Management (identifying ideal locations for new or expanded truck parking facilities)
- Level of Service Impacts, especially on truck travel time and travel time reliability
Examples of Freight Models
A list of real-world freight models can be found here.