We discuss the implementation of a rubber-banding function in an aggregate tour-based modeling framework. The aim of rubber-banding is to make the choice of destinations along an activity chain more realistic.
The idea behind it is that the destination choice for all activities is made based on a main activity, which allows us to determine more consistent stop locations along a tour. We show the implementation of rubber-banding in a generic matrix based format as part of the combined distribution/mode choice and destination choice process.
In a matrix based formulation for activity legs, the tour is divided into half-tours with a primary activity and an anchor location or home location. The destination choice for the main activity is calculated first. The location of stops along the half-tour is then calculated based on a composite impedance function. With this additional dependency, the choice of an intermediate stop between home and main activity is made such that the location of stops along the tours is more realistic. To make the process more generic and easier to calibrate, a weighting factor is used to specify how strictly the destination choice adheres to the proximity of the half-tour.
In addition to discussing the specifics of the methodology, a case study of the above implementation on a real world network will also be presented in order to add a practical context to the above theoretical formulation.