Validating the results of a route choice simulator
In calibration of any transportation model, some of the more typical measures used are volumes, speeds, travel time, and congestion (length and duration of queuing).These measures are used in a depends on more than link volumes.Model calibration objectives must be consistent with project objectives.Type of project, stakeholder objectives, and model considerations feed into the calibration objectives, as depicted in Figure 8.2.It is important that the trip characteristics (path and time of trip taken) reflect observed conditions and expectations.Increasingly, vehicle path data are becoming available through Bluetooth and GPS technologies, enhancing the ability to accurately assess tend to cover large areas, and the ability to obtain good calibration data for the entire model may be limited.The analysis team must understand the underlying simulation and modeling techniques in a model in order to understand how the model would react to network and demand changes and the measures of effectiveness that should be the focus of calibration.
This process also includes the statistical verification of the model outputs vis à vis the field-measured conditions.
Validation in transportation modeling is often used as the term for verification of the model outputs.
Validation in this guidebook is defined as a process in which a calibrated model is tested using a different set of existing traffic data to determine if the calibration parameters are applicable to other conditions.
The performance measures for calibration should also be consistent with the performance measures to be used to evaluate scenarios.
The analysis team should begin by identifying a set of “must-have” performance measures, then estimate the resource requirements, then gradually include additional “good-to-have” performance measures if budget and schedule allow.
The calibration acceptance targets may need to be adjusted based on the variability observed in the data.