Approximate Continuous K Nearest Neighbor Queries for Continuous Moving Objects with Pre-defined Paths

Abstract

Continuous K nearest neighbor queries (C-KNN) on moving objects retrieve the K nearest neighbors of all points along a query trajectory. In existing methods, the cost of retrieving the exact C-KNN data set is expensive, particularly in highly dynamic spatio-temporal applications. The cost includes the location updates of the moving objects when the velocities change over time and the number of continuous KNN queries posed by the moving object to the server. In some applications (e.g., finding my nearest taxies while I am moving), obtaining the perfect result set is not necessary. For such applications, we introduce a novel technique, AC-KNN, that approximates the results of the classic C-KNN algorithm, but with efficient updates and while still retaining a competitive accuracy. We evaluate the AC-KNN technique through simulations and compare it with a traditional approach. Experimental results are presented showing the utility of our new approach.

This research has been funded in part by NSF grants EEC-9529152 (IMSC ERC), MRI-0321377, and CMS-0219463, and unrestricted cash/equipment gifts from the Lord Foundation, Hewlett-Packard, Intel, Sun Microsystems and Raptor Networks Technology.