SPATIOGTEMPORAL DATAWAREHOUSES FOR TRAJECTORY EXPLOITATION
Project Number: P19
Project Manager: Dr. ir. Maurice van Keulen / Dr. Andreas Wombacher
Faculty of Electrical Engineering, Mathematics and Computer Science
Tel.: +31-53-4893688 / 4893772
Email: email@example.com / firstname.lastname@example.org
Internet technologies have created a fabric for information exchange far beyond the original intent to share multimedia information between humans. It has become a landscape where entities (people, sensors, devices, observatories) generate events bound in time and by geo- spatial locations, carving out event trajectories of interest. Collecting, archiving and annotation of these trajectories forms the basis for knowledge extraction, e.g., for fleet optimization to reduce traffic congestion and energy consumption, manage social communities, e.g., to find similar cases quickly for emergency handling (spam, DDOS), and to organize the community by mutual interest, e.g., holiday entertainment. Likewise, large-scale trajectory collections become the prime target for scientific discovery and natural disaster management, e.g., seismic and remote sensing.
In scientific terms, the above calls for major technological advances in the design, development and deployment of a spatial-temporal datawarehouse. The technical challenges are focussed on scale and responsiveness in the face of massive observations. One of the salient outcomes of this project is an open-source reference platform, called the Time-Trail Vault, which supports experimentation with datamining algorithms over trajectories, query processing over scalable databases derived from real-life trajectory use-cases, and community-centric trajectory tagging and profiling.
This project addresses the fundamental challenges emerging from databases populated with trajectory data using novel techniques and scientific insights, such as automatically evolving distributed database architectures and self-tuning database techniques to cope with the ever growing data volumes. Scaling towards the data volume and breadth of applications requires major innovations in the hitherto solutions developed for datawarehouses. It is a key topic in the major database conferences, e.g., VLDB, SIGMOD, ICDE and EDBT. MDL-based pattern mining such as forms the centre of a wide variety of algorithms and applications that all depend on models consisting of small sets of patterns. Generalizing these techniques to spatiotemporal and sensor data streams at a scale exemplified by the use cases is an open research topic at conferences such as KDD and PKDD. The realization of dataspaces around volunteered information requires (geospatial) entity recognition implicit handling of the uncertainty around vague objects and disambiguated conflicts, and user- guided data cleaning.
Project duration: 1-5-2011 / 1-5-2015
Project budget: 5.96 M-€ / 2.24 M-€ funding
Number of person/months: not yet known
Project Coordinator: CWI
Participants: CWI, University of Utrecht, UT, TomTom, KNMI, Hyves, EuroCottage, Arcadis, MonetDB
Project budget CTIT: 1.2 M-€ / 559 k-€ funding
Number of person/months CTIT: not yet known
Involved groups: Databases (DB)