On the generation of spatiotemporal datasets

Abstract

An efficient benchmarking environment for spatiotemporal access methods should at least include modules for generating synthetic datasets, storing datasets (real datasets included), collecting and running access structures, and visualizing experimental results. Focusing on the dataset repository module, a collection of synthetic data that would simulate a variety of real life scenarios is required. Several algorithms have been implemented in the past to generate static spatial (point or rectangular) data, for instance, following a predefined distribution in the workspace. However, by introducing motion, and thus temporal evolution in spatial object definition, generating synthetic data tends to be a complex problem. In this paper, we discuss the parameters to be considered by a generator for such type of data, propose an algorithm, called “Generate_Spatio_Temporal_Data” (GSTD), which generates sets of moving point or rectangular data that follow an extended set of distributions. Some actual generated datasets are also presented. The GSTD source code and several illustrative examples are currently available to all researchers through the Internet.