Abstract
With the rapidly growing availability of vehicle trajectory data, travel costs such as travel time and fuel consumption can be captured accurately as distributions (e.g., travel time distributions) instead of deterministic values (e.g., average travel times). We study a new path finding problem in uncertain road networks, where paths have travel cost distributions. Given a source and a destination, we find optimal, non-dominated paths connecting the source and the destination, where the optimality is defined in terms of the stochastic dominance among cost distributions of paths. We first design an A based framework that utilizes the uncertain graph to obtain the most accurate cost distributions while finding the candidate paths. Next, we propose a three-stage dominance examination method that employs extreme values in each candidate path's cost distribution for early detection of dominated paths, thus reducing the need for expensive distributions convolutions. We conduct extensive experiments using real world road network and trajectory data. The results show that our algorithm outperforms baseline algorithms by up to two orders of magnitude in terms of query response time while achieving the most accurate results.