Adjacency-based indexing for moving objects in spatial databases
thesisposted on 23.02.2017, 02:38 by Alamri, Sultan Mofareh
With the currently available positioning devices such as Global Positioning System (GPS), RFID, Bluetooth and WI-FI, the locations of moving objects constitute an important foundation for a variety of applications. However, with the recent developments new challenges are presented to traditional database technology. In traditional database systems, the data structures do not support the high-update objects since these structures have not been designed to accommodate dynamic updates of moving objects. Therefore, much research has gone into the outdoor moving objects (e.g. indexing and querying). However, moving objects in indoor spaces no less important than in outdoor. Indoor space refers to cellular spaces where objects are located based on their cells/rooms. Similarly, in some outdoor topographical spaces, moving objects applications focus only on the region/ cell where mobile/moving objects are located (not the exact coordinate location). Therefore, this thesis addresses three important basic issues in moving objects databases: (i) understating the variety of the moving objects' features and queries, (ii) adjacency/cellular indexing for moving objects in indoor space, and (iii) adjacency indexing for moving objects in outdoor space. This thesis starts by presenting a new taxonomy for moving object queries. This taxonomy provides a better understanding of the moving objects' features and their variety of queries. The taxonomy for moving object queries is based on five perspectives. First is the Location perspective, which includes common spatial queries such as K nearest neighbours (KNN), range queries and others. Second is the Motion perspective, which covers direction, velocity, distance and displacement queries. Third is the Object perspective, which includes the type status queries and the form status queries. Fourth is the Temporal perspective which includes many queries such as the trajectory, timestamped, and period queries. Last is the Patterns perspective, which includes many patterns such as spatial movement patterns and temporal movement patterns. Each perspective is explained with illustrated examples. In addition, in indoor spaces, we propose a new cells adjacency-based index structure for moving objects. The new index structure focuses on the moving objects based on the notion of cellular space. Therefore, an indoor filling space algorithm is proposed to represent overlapping between the cells. Then, the index structure for indoor space uses the indoor filling space algorithm for efficient adjacency grouping and serving the indoor spatial queries. The temporal side has been presented using three different techniques: Trajectories Indoor-tree (TI-tree), Moving Objects Timestamping-tree in indoor cellular space (MOT-tree) and Indoor Trajectories Deltas index based on the connectivity of cellular space (ITD-tree). Moreover, in order to improve the performance of the index structure in indoor space, the density of the moving objects needs to be considered. Thus, this thesis presents a new index for moving objects (Indoor(d)-tree) that distinguishes between the high and low density cells. Moreover, it presents a new index structure for moving objects in multi-floor indoor environments (Graph-based Multidimensional Indoor-Tree or GMI-tree) which uses the indoor connectivity graph to group the moving objects multidimensionally and to support wings/sections positioning queries. Moreover, in outdoor spaces, we extended the regional/adjacency indexing in indoor spaces to be applicable in topographical outdoor spaces. Thus, we propose an efficient data structure index (Topographical Outdoor-tree or TO-tree) for the moving objects based on their cellular location. The proposed Adjacent Level Algorithm and Connection Cells Algorithm are used to determine the connectivity level of each cell in the topographical outdoor space. Therefore, the TO-tree will group the moving objects based on their adjacency, thereby reducing the update costs and efficiently supporting the spatial queries and adjacency queries. Furthermore, based on our taxonomy, a very limited number of data structures concentrates on the motion vectors in the construction of the moving objects' data. Therefore, we propose a new index structure (DV-TPR*-tree) that supports velocity and direction queries, beside the common spatial queries. To sum up, this thesis introduces a new level of indexing called adjacency indexing which proves to be efficient and robust in indoor and outdoor spaces.