Modellierung navigatorischer Fahrerpräferenzen für die personalisierte Routenberechnung

The routing component of nowadays navigation devices bases on a shortest path calculation using an objective function and edge attributes of an underlying graph. Minimizing the route duration or route length are the most common objective functions to meet drivers’ general navigation preferences for a route.

In this work, we include the navigation preferences of different types of drivers in the path calculation by personalizing the routing component. The objective function as well as edge attributes are adapted to drivers’ preferences in an automated way. To describe drivers’ navigation preferences in a navigation context, it is necessary to model driver behavior appropriately. Three different models are developed by analyzing different perspectives of drivers’ navigation behavior covering the dimension of time as well as the dimension of space.

The first, time-oriented model examines the observed speeds on the edges. For the second, space-oriented model, a measure considering the entirety of edges is provided as to characterize the coverage of a driver’s mobility network. The third model analyses routes consisting of paths of edges considering both dimensions. Correlation analyses show that results obtained from the different models confirm each other. Based on the results, a well-founded two-dimensional model is built from well-known attributes and to classify the drivers in respect to their navigation preferences.

This model is used for personalizing the objective function and the edge attributes separately and in a combined way. The calculated routes are compared with a standard route in respect to different measures to evaluate the personalization. All steps of modelling and personalization are implemented for a large scale data set. In this way, the improvement of the navigation device by a model-based personalization is quantified for each driver as well as for the determined classes of the sample.