Notification agents for mobile route guidance

Ubiquitous and mobile computing has become an important field in computer science. The provision, management and transmission of information within computer systems and for human-computer interaction is required for many applications and the individual needs of the mobile user are manifold. Due to information overflow it is difficult for the user to access specific pieces of information while key information for decision makers is often missing. Technologies for the emerging field of information logistics are studied in order to enable an optimal information supply with respect to the individual users needs. A dominant application area for this is mobile route guidance, especially when it comes to messaging about alternative routes. Such alternative routes might be advantageous in case of unexpected incidents such as the emergence of an additional train (or cancellation) so that the traveller may reach his/her destination earlier. The main contribution of this thesis is the conceptualization of a method for the selection of individual notifications for mobile route guidance which considers future notification options and the information quality gain to be achieved. Two scenarios are used for the demonstration of continuous individual messages for the traveller notifying him about the best route to follow. Explicit models for the uncertainty about future situations are introduced and the uncertainty reduction by waiting for more information is quantified. The topology of the transportation network is considered and used as a structural model for the construction of influence diagrams as decision models. These influence diagrams answer the questions about the best notifications to be sent both with respect to time and content. Our approach goes beyond probabilistic route planning, which selects a route with maximum expected utility. Other than that, future notification options are considered and the best notification policy is selected rather than the best route. By this, the expected utility can be improved with respect to the standard approach.