Abstract
In this paper we are introducing a self learning tool for travel time estimation in signalized urban networks based on probe data. The main feature of this tool is, that it can be applied with a basic network description instead of a detailed modeling of the network structure. We show how probe data can be utilized to train a Bayesian network to forecast the travel time on a route along an arterial road. In the conclusion we take a critical look on the limitations of such a system and possible extensions to increase its performance.