An efficient and scalable method for aggregate nearest neighbor queries on time-dependent road networks. Issue 105 (March 2022)

• Record Type: Journal Article
• Authors: Ma, Hui; Tang, Yong
• Is Part Of: Information systems. Issue 105(2022)
• Publication Date: 2022-03
• Subjects: Time-dependent road network -- Aggregate nearest neighbor -- Graph query -- Algorithm; Database management -- Periodicals; Electronic data processing -- Periodicals; Bases de données -- Gestion -- Périodiques; Informatique -- Périodiques; Database management; Electronic data processing; Periodicals; 005.7
• DOI: 10.1016/j.is.2021.101925
• Languages: English
• ISSNs: 0306-4379

Abstract

Abstract: We study the k aggregate nearest neighbor queries on time-dependent road networks (TDRNs), assuming that the travel time to traverse an edge depends on the time it is initiated. Given a set of query points, a set of points of interest (POIs), an aggregate function (e.g., max ) and a number k, the query returns k POIs p in increasing order of the values of the aggregate travel times from the query points to p . We investigate two forms of queries. The single departure time query requires all the query points to simultaneously start at a fixed departure time given as a query condition. In addition, we propose a more flexible form of query, the time interval query that allows the query points to independently depart at any arbitrary time in a given departure time interval. Existing methods are either too inefficient or sensitive to query factors, i.e., the query running time is greatly affected by some factors like the number or the distributions of query points and POIs. We propose an efficient and scalable method to answer the two forms of queries using time-dependent G-Tree (TDGtree). TDGtree is a tree-structured index that could efficiently return the minimum travel time between two locations in a TDRN. We devise a function to evaluate which regions are likely to contain the promising POIs, and we employ a minimum-first-search strategy to preferentially visit the promising regions so that we could efficiently pinpoint the k optimal POIs. Experiments by varying a number of factors demonstrate the superiority of our method. In terms of query efficiency, our method achieves query speed faster than the states-of-the-arts up to two orders of magnitude. In the aspect of scalability, our method has relatively steady query running time regardless of the query parameter value changes. Highlights: Traditional ANN query on time-dependent road network is strict in application. A novel and flexible interval departure time query is formally defined. Fast and scalable single/interval departure time query algorithms are presented. Experiments were conducted to confirm the performance on real/synthetic datasets.