Date: 4th of November;   Location: Delft,  D-CIS Lab 

Contemporary net-centric applications, such as large scale crisis management, maritime security, disease and pollution monitoring, etc., can greatly benefit from the fusion of large amounts of heterogeneous information stemming from static and mobile sensors as well as intelligence obtained from human observers, databases and the World Wide Web. By combining multiple information modalities, the content of the relevant information in the fusion system is increased, which in turn can mitigate the impact of noise and reduce ambiguities.  In fact, many contemporary fusion challenges can be solved only by combining heterogeneous types of information. 

However, the heterogeneity of the available information and dynamic constellations of information sources, often dispersed throughout systems of mobile platforms, introduce substantial challenges which require novel approaches. Often it is not trivial to build systems which allow sound correlation of information of different types and quality. This typically requires advanced approaches to modeling and processing. In addition, real world systems are likely to require large models whose construction and maintenance can become intractable, if not addressed in a systematic manner. Modularization of fusion processes, the use of Service Oriented Architectures and collaborative design methods might provide solutions to these challenges. This in turn requires design methods and tools which exploit inherent properties of the used modeling and processing techniques. Moreover, a huge body of potentially useful information can be found in easily accessible electronic documents and on the Internet. However, such unstructured information can be used by automated fusion processes only if it is appropriately preprocessed and transformed. This is a non-trivial problem which requires suitable techniques, such as semantic networks, pattern recognition, etc.

The main goals of the workshop are (i) identification of major challenges in contemporary fusion applications and (ii) discussion on methods supporting efficient and sound implementation of advanced information fusion systems. The discussion will be guided by distinguished experts from industry and academia, who will address the following relevant topics:

• Identification of challenges (modeling, inference, pattern recognition, semantic processing, net-centric operation).
• Understanding the problem complexity and tractable implementation of reliable fusion systems.
• Sound information processing. What does this mean and why is it important?
• Contemporary modeling methods and algorithms which can cope with heterogeneous information and facilitate distribution of fusion processes.
• Methods facilitating semantic processing.
• Pattern recognition.

Invited speakers:
Subrata Das (Xerox Research Centre Europe, Grenoble)
Hans Driessen (Thales Nederland B.V.)
Bert Kappen (Radboud Universiteit, Nijmegen)
Jean Rohmer (Universite Leonard de Vinci, Paris)

Program:
10:00 Welcome and Introduction  

10:10 Selected fusion applications (Hans Driessen, Subrata Das, Jean Rohmer, Gregor Pavlin)

11:15 Coffee break

11:30 Stochastic optimal control, active sensing and learning. (Bert Kappen)

12:30 Lunch

13:30 Quest for structured information (Subrata Das)

14:30 Coffee Break

14:45 “Literatus Calculus”, a proposal towards information fusion for situation assessment (Jean Rohmer)

15:15 Efficient and robust fusion of heterogeneous information using causal Bayesian networks. (Gregor Pavlin)

15:45 Coffee Break

16:00 Round table (discussion on relevant research directions)

17:00 Drinks

Abstracts

Please download the pdf for the abstracts