Private Transport Analysis

Analyzing transportation networks, primarily private transport, can be done at three resolutions:

• Microscopic:

  At this level, simplified mathematical models of driving behavior control the movement of individual vehicles, their interactions with other vehicles and objects along the lane, during lane changes, and at intersections. Road capacity emerges as an output of these interactions. Due to the high level of detail, a degree of stochasticity is incorporated, and the aggregated results from multiple runs are considered reliable.

• Mesoscopic:

  This resolution encompasses a spectrum of models that bridge the gap between microscopic and macroscopic approaches. Generally, these models leverage advantages from both classes to model large networks with greater detail in less time. Consequently, some mesoscopic models lean closer to microscopic models, while others are more akin to macroscopic ones.

• Macroscopic:

  In this class, a simplified mathematical function, known as a volume-delay function (VDF), represents the relationship between traffic volume and delay on each link, node, and turn within the network. Here, capacity is an input to the VDF.

Traffic volumes in network models can be determined using various methods:

• Fixed turning shares across scenarios:

  This method is suitable when a project is not expected to alter route choices. The shares can be directly derived from traffic counts. These shares can vary over time of day.

• Estimated future turning shares:

  With this method, future turning shares are estimated based on available data, scenario, and assumptions, or are obtained from a large-scale model. This approach is well-suited for most traffic impact analysis (TIA) studies.

• Traffic assignment:

  In this method, vehicles select their paths based on a choice function. The calibration model is first validated, then applied to future conditions. This method is appropriate for large-scale studies, and the remainder of this discussion focuses on such models.

To build a traffic assignment model, road network data is acquired from online sources and/or the network of a metropolitan planning organization (MPO) model. The latter is usually too coarse (lacks sufficient detail) for traffic studies. Signal data can be sourced from signal operators and online data providers.

To estimate travel demand for each peak period, data from MPO models and existing traffic counts are utilized, or new traffic counts are collected. Online transportation data sources are also valuable. Demand estimation for traffic projects is a complex, iterative process requiring extensive checks and validations, as the quality of model calibration heavily depends on demand accuracy. This is even more critical in dynamic models.

Depending on the intended model outputs, either static or dynamic assignment is applied:

• Static assignment:

  This method does not have a time dimension, and outputs represent either averages or totals (depending on the output type) over the analysis period.

• Dynamic traffic assignment (DTA):

  In this method, the time dimension is accounted for, and results are generated for each time interval. Depending on the specific method, DTA considers departure times, travel times, flow propagation, and spillbacks within the network.

Based on the assignment method and model resolution, different parameters are considered when calculating delays through roads (links) and at intersections (nodes). These range from VDFs and flow-density fundamental diagrams to HCM-based methods, as well as mesoscopic and microscopic models.

The calibrated model can be used to analyze various road network alternatives, conducting subnetwork analysis (e.g., with micro-simulation), and performing network-wide signal timing optimization and coordination. The outputs of such models include, for example, vehicle miles traveled (VMT), vehicle hours traveled (VHT), delay, and queue length.

Due to the uncertainties associated with the market penetration of autonomous vehicles (AVs), a scenario-based approach is employed to evaluate future conditions, based on different assumed market penetration rates for AVs.