Daniel Delling and Renato Werneck
We present a unified framework for dealing with exact point-of-interest (POI) queries in dynamic continental road networks within interactive applications. We show that partition-based algorithms developed for point-to-point shortest path computations can be naturally extended to handle augmented queries such as finding the closest restaurant or the best post office to stop on the way home, always ranking POIs according to a user-defined cost function. Our solution allows different trade-offs between indexing effort (time and space) and query time. Our most flexible variant allows the road network to change frequently (to account for traffic information or personalized cost functions) and the set of POIs to be specified at query time. Even in this fully dynamic scenario, our solution is fast enough for interactive applications on continental road networks.
Publisher Microsoft Technical Report