Abstract
Aggregate nearest neighbor queries return a point with a minimum net distance from a set of query points. Consider, for example, group of friends located at specific locations (query points) that want to meet at a restaurant (a point) such that they travel the minimum sum of distances in order to meet. In this paper, we proposed a fast algorithm, A 2 N 2 , to answer such aggregate nearest neighbor queries on road networks based on pre-computation. An assortment of optimized data structures and techniques are used so as to reduce the overall computation time. Additionally, by focusing on reducing the amount of pre-computed data stored and using efficient ways to retrieve and use them during query time, the algorithm is computationally faster at the cost of being minimally approximate. Experiments on real road network data sets demonstrate the impact of input parameters on the query processing time and supports the claim. It was observed that the pre-computation time and query processing time for A 2 N 2 was respectively in the orders of up to 1000 and 100 times faster than that of a Voronoi based ANN approach. The minimum normalized path distance deviation across all data sets for A 2 N 2 was only 2% with the computed path distances comparable to a Voronoi based approach.