Establishment of Center of Excellence in Multisource Information Fusion
Final rept. Dec 1996-Dec 2000
CALSPAN UB RESEARCH CENTER BUFFALO NY
Pagination or Media Count:
This Final Technical Report is a summary of the work conducted on the following research tasks 1 Correlation Science, 2 Learning in Distributed Fusion Systems for Decision Making and 3 Information Sharing in Distributed Fusion Applications. The Correlation Science task has derived a new association scoring approach, JHSO, and compared its performance with four traditional association approaches. The JHSO yields the best association and estimation results for problems where an accurate report and track are being considered for association versus another less accurate track e.g., from off-board. The other approaches are biased towards less accurate tracks since they do not consider the resultant state accuracy in their scoring for alternative associations. The Learning in Distributed Fusion Systems for Decision Making task addressed issues of cooperative group learning from examples in multi agent distributed system for decision making. An approach to both individual learning with decision integration and distributed learning utilizing the Dempster Shafer theory of evidence has been introduced. More research is needed to address fundamental issues of the problem of cooperative learning in problem solving systems such as utilization of a prior knowledge, incorporation of symbolic and numeric information, and convergence of the process. The Information Sharing in Distributed Fusion Applications task presented a model and testbed for situation assessment. Many parameters affecting information sharing and strategies for information sharing were considered but a few key parameters are still missing. A trade off of 15 detection rate for a 30 saving in communication was shown. Cost and other factors such as models of other agents need to be included in the testhed to make it more useful.
- Information Science