Travel Time Prediction Using Floating Car Data Applied to Logistics Planning

Travel time information plays an important role in transportation and logistics. Much research has been done in the field of travel time prediction in local areas, aiming at accurate short-term predictions based on the current traffic situation and historical data of the area. In contrast, literature on prediction methods for long-range trips in large areas is rare, although it is highly relevant for logistics companies to manage their fleet of vehicles. In this paper, we present a new algorithm for predicting the remaining travel times of long-range trips. It makes use of nonparametric distribution-free regression models, which are applicable only in the presence of a sufficiently large database. Since, in contrast to local areas, such a base is visionary for large areas, we bring into play a dynamic data preparation to artificially enlarge the database. The algorithm also takes into account that routes of long-range trips are not completely given in advance but are rather unknown and subject to change. We illustrate our algorithm by means of simulations and a real-life case study at a German logistics company. The latter shows that, by our algorithm, the average relative error can be halved compared with conventional methods.