Creating Moving Objects Representations for Spatiotemporal Databases

Background

The Database Management Systems (DBMS) provide standard methods for storing, managing and querying data efficiently. They have been initially tailored for handling one-dimensional data, such as, strings, dates or numbers, but the success of the relational model and the widespread use of DBMS in many organizations have led to the development of new solutions for dealing with more complex data types. This is the case of spatial DBMS extensions that focus on the representation of geographical objects or phenomena, which can be modeled using abstractions such as points, lines or polygons. Complexity increases because spatial data is usually represented in two or higher dimensional spaces and the topology of spatial entities need to be considered as well. Oracle Spatial and PostGIS are notable examples of spatial DBMS extensions currently in use.

In addition, spatial data may also be dynamic and the spatiotemporal behavior of geographical objects or entities across time is essential for several applications. This has gained particular importance with recent developments in remote sensing and communication technologies. One of the most challenging topic is the representation of continuously changing spatial data about moving objects with extent, hereafter referred to simply as moving objects.

The abstract definition of a moving object is a triple where is a temporal value, is the moving object’s geometry at a certain time and is a continuous function defining the transformation of during (Chomicki, Haesevoets, Kuijpers, & Revesz, 2003). The semantics is,where and are coordinates, is a time instant and is a continuous function that describes how the object moves or changes during time. This is an abstract representation of moving objects where the spatial and the temporal values are defined as infinite set of points, which needs to be translated into discrete (finite) data models suitable for implementation in databases. The main spatiotemporal data models were based on constraints (Grumbach, Rigaux, & Segoufin, 2001) and Abstract Data Types (ADT) (Güting, et al., 2000). Subsequent works have mainly followed the ADT approach because this data model can be smoothly built into extensible DBMS such as the object-relational DBMS.