Information Acquisition in Data Fusion Systems

By purposefully utilising sensors, for instance by a data fusion system, the state of some system-relevant environment might be adequately assessed to support decision-making. The ever increasing access to sensors offers great opportunities, but also incurs grave challenges. As a result of managing multiple sensors one can, e.g., expect to achieve a more comprehensive, resolved, certain and more frequently updated assessment of the environment than would be possible otherwise. Challenges include data association, treatment of conflicting information and strategies for sensor coordination.

We use the term information acquisition to denote the skill of a data fusion system to actively acquire information. The aim of this thesis is to instructively situate that skill in a general context, explore and classify related research, and highlight key issues and possible future work. It is our hope that this thesis will facilitate communication, understanding and future efforts for information acquisition.

The previously mentioned trend towards utilisation of large sets of sensors makes us especially interested in large-scale information acquisition, i.e., acquisition using many and possibly spatially distributed and heterogeneous sensors.

Information acquisition is a general concept that emerges in many different fields of research. In this thesis, we survey literature from, e.g., agent theory, robotics and sensor management. We, furthermore, suggest a taxonomy of the literature that highlights relevant aspects of information acquisition.

We describe a function, perception management (akin to sensor management), which realizes information acquisition in the data fusion process and pertinent properties of its external stimuli, sensing resources, and system environment.

An example of perception management is also presented. The task is that of managing a set of mobile sensors that jointly track some mobile targets. The game theoretic algorithm suggested for distributing the targets among the sensors prove to be more robust to sensor failure than a measurement accuracy optimal reference algorithm.