Spatio-Temporal Indexing: Current Scenario, Challenges and Approaches

1. MOTIVATION With rapid advancements in computing hardware, tracking devices such as GPS receivers and sensors have become pervasive, generating a large amount of spatio-temporal data, such as measurements of temperature, pressure, air quality, traffic, etc. using sensors, GPS data from mobile phones and data from radars that capture location information about people and other moving objects such as cars and aeroplanes. This has enabled a wide variety of spatio-temporal applications, resulting in a renewed interest in techniques for handling spatio-temporal data. Over the past two decades or so, a large number of indexes for supporting spatial, temporal and spatio-temporal data have been independently proposed in the database and data mining communities. However, there exists no clear-cut guidelines or a prescriptive formula for pointing out which index should be chosen when specific needs of the underlying application are known. In addition, since spatiotemporal indexes have been proposed under various domains, it is hard for researchers and practitioners to determine whether some specified indexes are indeed available to address the problem at hand. For instance, an index like PO-Tree [7] is suitable for monitoring static spatio-temporal objects (such as sensors, cell-phone towers, etc.) but it is completely undesirable for handling moving object data (e.g., location tracking of cell-phone users, GPS tracking of vehicles and so on). Likewise, if the semantics of the application require indexing trajectories of moving objects, only a specific set of indexes (such as PA-Tree [6]) are useful whereas others such as (APR-Tree [3]) are undesirable. We design this tutorial to expose the audience to the vast reservoir of spatio-temporal indexing techniques that are available in literature. In addition, apart from introducing the various indexes, our aim is to analyze the pros and cons of different indexing mechanisms when applied to various diverse scenarios. Given the large recent interest in spatio-temporal data analytics among corporates and academia, we hope that this tutorial is well-positioned in time to enhance and enrich the understanding of spatio-temporal data processing. Further, we think that the subject matter of this tutorial is a perfect fit for the COMAD conference that has a focused track for data management and its disciplines.