Succint and self-indexed data structures for the exploitation and representation of moving objects

Abstract

This thesis deals with the efficient representation and exploitation of trajectories of objects that move in space without any type of restriction (airplanes, birds, boats, etc.). Currently, this is a very relevant problem due to the proliferation of GPS devices, which makes it possible to collect a large number of trajectories. However, until now there is no efficient way to properly store and exploit them. In this thesis, we propose eight structures that meet two fundamental objectives. First, they are capable of storing space-time data, describing the trajectories, in a reduced space, so that their exploitation takes advantage of the memory hierarchy. Second, those structures allow exploiting the information by object queries, given an object, they retrieve the position or trajectory of that object along that time; or space-time range queries, given a region of space and a time interval, the objects that are within the region at that time are obtained. It should be noted that state-of-the-art solutions are only capable of efficiently answering one of the two types of queries. All of these data structures have a common nexus, they all use two elements: snapshots and logs. Each snapshot works as a spatial index that periodically indexes the absolute position of each object or the Minimum Bounding Rectangle (MBR) of its trajectory. They serve to speed up the spatio-temporal range queries. We have implemented two types of snapshots: based on k2-trees or R-trees. With respect to the log, it represents the trajectory (sequence of movements) of each object. It is the main element of the structures, and facilitates the resolution of object and spatio-temporal range queries. Four strategies have been implemented to represent the log in a compressed form: ScdcCT, GraCT, ContaCT and RCT. With the combination of these two elements we build eight different structures for the representation of trajectories. All of them have been implemented and evaluated experimentally, showing that they reduce the space required by traditional methods by up to two orders of magnitude. Furthermore, they are all competitive in solving object queries as well as spatial-temporal ones.