Synthesizing Datasets for Pervasive Spaces

Persistent problems in obtaining test data while developing technologies for pervasive spaces include the lack of data generation and representation standards, limited amount of available test data, incompleteness of existing data sets, and high cost of implementing a given pervasive space. This not only hinders the development and design of pervasive spaces, but also decreases the ability of researchers to compare results with those of other areas in computer science. For example, in the interaction between sensors and actuators, testing the robustness of new algorithms and evaluating risk requires large datasets that portray different scenarios. However, associated costs can be prohibitive as the data burden increases. In response to this situation, we propose a Pervasive Space Simulation Technique (PSST) that uses Markov chains, non-homogenous Poisson processes, and distribution fitting for generation of synthetic test data. PSST provides a realistic simulation of events in the pervasive space.