Representative dissimilar path queries: accommodating human movement dynamics in road networks

We introduce a representative dissimilar path (RDP) query, a novel type of path query in road networks. The k representative paths (RPs) between a source and a destination locations have k smallest costs for a feature (e.g., length, number of road intersections, or straightness). Given x features and k, an RDP query returns a set of paths for a source-destination pair such that the path set includes at least one of the k RPs for every feature, and the path set's similarity score is minimized. We formulate a novel measure to quantify the similarity of a set of paths. Considering different road features and incorporating the novel similarity measure in the computation of RDPs allow us to accommodate the human movement dynamics between two locations in an effective way. Finding the RDPs is a computational challenge because an RDP query requires computing the RPs for multiple features and then finding the RDPs from an exponential number of path combinations. We develop an efficient solution to answer RDP queries. The underlying ideas behind the efficiency of our algorithms are the refinement of the search space, finding the RPs for multiple features with a single search, and exploiting both the lower and upper bounds of the path set's similarity score while identifying the RDPs. We show the efficacy of the RDP query and the efficiency of our solution to answer the RDP query in extensive experiments using real datasets.