This project contributes to the exciting field of geo-stream processing. Specifically, an integrated real-time geo-stream processing and monitoring system called AegisDB will be built and the research environment and results will be leveraged to stimulate learning at the K-12, undergraduate, and graduate levels.
The AegisDB system includes a Geo-Stream Algebra which uses a data-type- based approach as opposed to a traditional tuple-based approach for representing and querying geo-streams. The STREAM data types of the algebra unify static data, traditional streams, geo-streams from fixed locations, and geo-streams that move. The proposed Aggregate Algebra generalizes the operator GROUP BY to generalized aggregations (GAs) and bridges the fundamental gap between point observations from sensors and spatio-temporally continuous phenomena. Several novel spatial-centric operator optimization techniques are proposed, which include a velocity-based filtering that utilizes the physical limitations of most moving objects and a spatial-coverage-based query combination that combines geo-stream queries based on their intended spatial coverage.
The proposed AegisDB system will transform the ways to define and monitor more sophisticated spatio-temporal events for alerting purposes. The result is the significantly improved realtime access to important spatio-temporal events related to our lives by professionals and the general public. The system will be critical for important domains including transportation, environmental science, hazard monitoring, and emergency responses.
This project has strong education and outreach components. Environmental models for K-12 teachers and students using real-time environmental datasets will be designed. An informal session on how to collect, store, and distribute environmental data will be presented through the Family Fun Science Saturday events of the Elm Fork Education Center of UNT. A course module on geo-stream processing will be developed. Women and minorities will be recruited into the research project. Undergraduate students looking for research opportunities will be mentored.
For further information concerning this project see the project web page: URL: www.cse.unt.edu/~huangyan/AegisDB/
The goal of this project is to build a real-time geo-stream processing and monitoring system that: (1) uniform geo-streams and large volumes of static infrastructure datasets (e.g., road networks), (2) allows spatially explicit monitoring queries, and (3) spits out results in real time even in disastrous situations when data volumes are often high. Such a system can serve as the backend of many spatial reference stream systems such as smart transportation systems and a sensor web just as traditional databases did for many applications. Our first major research contribution is the design and implementation of a geo-stream algebra to support moving and evolving spatial objects, e.g. trajectories and evolving regions, in real-time. Our research activity focused on designing a data-type-based approach where an entity’s spatial extent over time can be managed as a data type and participate in common spatio-temporal predicates. The geo-stream algebra has a set of operators that are simple, closed, expressive, and easy to implement to support typical geo-stream queries. Our second major research contribution is the design and implementation of an aggregate algebra that generalizes GROUP BY query. This allows spatial cluster queries to be posted which are important for many applications but were not supported before. Our third major research contribution is operator scheduling and query optimization to support fast and efficient queries and analytics on large volume of geo-streams in real-time. Throughout the project, many students have been trained in the area of database systems, map matching, real-time scheduling, data stream processing, spatio-temporal data mining, and performance evaluation. Specifically, students have been trained in polygonizatoin algorithms, kinetic data structures, streaming data abstraction, geo-stream query languages, stream optimization, visualizing geo-streams, and big geo-stream data analytics. Throughout the project, 7 Ph.D. students have been trained and 3 Ph.D. theses have been developed and successfully defended. A dozen masters and undergraduate students have participated in the project. Four undergraduate students were supported and resulted in 5 peer-reviewed undergraduate co-authored papers in top tier journal and conferences. We hosted 2 projects for the Research Experiences for Teachers (RET) and 1 TechFest program for K-12 students.
