Spatio-Temporal Multi Data Stream Analysis with Applications in Team Sports (PhD Thesis, finished)
The amount of live data about individuals which can be collected is steadily growing. These days, humans can be equipped with physical devices or observed with cameras in order to capture information such as their positions, their health state, and the state of their environment. Fitness trackers and health applications which analyze the state and the behavior of an individual on the basis of the data that are captured for this individual are already widely used.
However, humans rarely act alone but rather collaborate in teams in order to achieve a common objective. For instance, football players collaborate to win a match and firefighters collaborate to extinguish a forest fire. Analyzing the collaborative team behavior on the basis of data about the individuals which form the team is not only interesting but further poses several challenges on the system that performs the analyses. The focus of this thesis is to address these challenges.
We define a data model and a system model in order to provide a theoretical basis for implementing a system that is suited to serve as a foundation for developing team collaboration analysis applications. Both models are novel with respect to the fact that they take the particularities of team collaboration analysis applications, such as the semantics of their input and output data, into account. Moreover, we establish a strong foundation for using the spatial and temporal information which play a central role in analyzing the collaborative behavior of a team. More precisely, we define basic spatial functions and relations and present an extensive stream time model which goes far beyond existing literature on stream time notions and comprises a novel simultaneousness concept.
After establishing the theoretical basis, we present StreamTeam, our generic real-time data stream analysis infrastructure which is designed to be used as a foundation for developing team collaboration analysis applications. The data stream analysis system at the heart of StreamTeam is a prototype implementation of our models which further introduces novel approaches to assist domain experts without a profound software engineering background in developing their own analyses. Moreover, we present StreamTeam-Football, a realtime football analysis application which is implemented on top of StreamTeam. StreamTeam-Football is the first analysis application which performs complex team behavior analyses in a football match in real-time, visualizes the live analysis results in a user interface, and stores them persistently for offline activities.
Prof. Dr. Peter Michael Fischer (University of Augsburg)
Prof. Dr. Heiko Schuldt (University of Basel)
Date of Defense
Funded since September 2017 by the Hasler Foundation in the Cyber-Human Sytems program (contract no. 16074).