[Back]

A. Schörgendorfer, W. Elmenreich:
"Extended Confidence-Weighted Averaging in Sensor Fusion";
Talk: Junior Scientist Conference, Vienna, Austria; 2006-04-19 - 2006-04-21; in: "Proceedings of the Junior Scientist Conference 2006", (2006), ISBN: 3-902463-05-8; 67 - 68.

Systems that deduce information about their environment from sensor data can increase the accuracy and reliability of such information by replicating sensors and combining or "fusing" their measurements. The corruption of single measurements through failures of the sensor or random noise can thus be masked. Research conducted in this field so far, however, has not taken into consideration that the errors committed by sensors are not always uncorrelated. Ignoring such dependencies can lead to a suboptimal fused result. In this work, the approach of confidence weighted averaging is extended to take such dependencies into consideration. The effect of the new approach on fusion results is analyzed using sensor information from a mobile robot.

http://aleph.ub.tuwien.ac.at/F?base=tuw01&func=find-c&ccl_term=AC06586984