Evaluating continuous K-nearest neighbor query on moving objects with uncertainty

摘要

Continuous K-nearest neighbor (CKNN) query is one of the most fundamental queries in the field of spatio-temporal databases. Given a time interval [ts,te], a CKNN query is to retrieve the K-nearest neighbors (KNNs) of a moving user at each time instant within [ts,te]. Existing methods for processing a CKNN query, however, assume that each object moves with a fixed direction and/or a fixed speed. In this paper, we relieve this assumption by allowing both the moving speed and the moving direction of each object to vary. This uncertainty on speed and direction of a moving object would increase the complexity of processing a CKNN query. We thoroughly analyze the involved issues incurred by this uncertainty and propose a continuous possible KNN (CPKNN) algorithm to effectively find the objects that could be the KNNs. These objects are termed the possible KNNs (PKNNs) in this paper. A probability-based model is designed accordingly to quantify the possibility of each PKNN being the KNN. In addition, we design a PKNN updating mechanism to rapidly evaluate the new query result when object updates occur. Comprehensive experiments are conducted to demonstrate the effectiveness and the efficiency of the proposed approach.