A Route Advice Agent that Models Driver Preferences

Generating satisfactory routes for driving is a challenging task because the desirability of a particular route depends on many factors and varies from driver to driver. Current route advice systems present a single route to the driver based on static evaluation criteria, with little or no recourse if the driver finds this solution unsatisfactory. In this paper, we propose a more flexible approach and its implementation in the Adaptive Route Advisor. Our route advice agent interacts with the driver to generate routes that he or she finds satisfactory, uses these interactions to build a model of the driver’s preferences, and then uses the model to generate better routes for that driver in future interactions. As the preference model becomes more accurate, the need for interaction decreases and the agent’s autonomy increases. We also present a pilot study on using route selections to construct a personalized model of driver preferences.