A new method to index and store spatio-temporal data

TL;DR

This paper introduces a novel compressed data structure based on $k^{2}$-trees for efficient storage and querying of spatio-temporal object trajectories, significantly reducing space while maintaining quick query responses.

Contribution

It presents a new method combining $k^{2}$-trees and relative movement encoding for compact, query-efficient spatio-temporal data storage, with an effective compression technique.

Findings

Significant space savings demonstrated in experiments

Efficient query response times achieved

Effective compression of relative movements

Abstract

We propose a data structure that stores, in a compressed way, object trajectories, which at the same time, allow to efficiently response queries without the need to decompress the data. We use a data structure, called $k^{2}$-tree, to store the full position of all objects at regular time intervals. For storing the positions of objects between two time instants represented with $k^{2}$-trees, we only encode the relative movements. In order to save space, those relative moments are encoded with only one integer, instead of two (x,y)-coordinates. Moreover, the resulting integers are further compressed with a technique that allows us to manipulate those movements directly in compressed form. In this paper, we show an experimental evaluation of this structure, which shows important savings in space and good response times.