Kalman-Tree: an Index Structure on Spatio-Temporal Data

In the research area of spatio-temporal databases, we need to answer the queries like ”select the objects which will be in a specific area at a future time” quickly and accurately. Solutions to this question include two aspects: a mathematic model to predict future location with adequate accuracy and an index structure that can retrieve the qualified objects quickly. Currently, the proposed future location estimation in this field is limited to some simple linear functions of velocity. It might not be able to get ideal prediction results as the prediction algorithms are too simple. This paper proposes a technique targeting the future location estimation problem in spatiotemporal databases. It uses the useful estimator Kalman filter to predict the continuous trajectory of the moving objects in multi-dimensional space. Based on the algorithm, we propose a modified R-tree indexing structure (KR-tree) that is well suitable and efficient for the Kalman equations.