Mahdi Khodabandeh

PhD Thesis Defense Session

Abstract:
Uncertainty evaluation for sensor fusion results as an open problem to justify the results in an efficient manner is concentrated in this thesis. The Aggregate uncertainty measure, AU, which is presented in Dempster-Shafer (DST) framework, is generalized in two procedures to be applicable to the fusion theories based on free and hybrid Dezert-Smarandache model. These fusion theories are Dezert-Smarandache Theory (DSmT) of combination including DSmC and DSmH also Proportional Conflict Redistribution (PCR) rules of combination. The new developed uncertainty measures are called Generalized Aggregate Uncertainty measure-1 (GAU1) and Generalized Aggregate Uncertainty measure-2 (GAU2). Requirements for these uncertainty measures are discussed theoretically. Additionally, modified forms of the AU measure, GAU1 measure and GAU2 measure are presented. Moreover, in order to achieve a well-organized decision making, a hierarchical sensor fusion framework with supervision of an uncertainty evaluator is introduced which it utilizes the uncertainty measures.
Capabilities of the developed uncertainty measures in order to evaluate the DSmT-based fusion results in the information fusion framework are verified by two experimental studies. The designed experiments are target differentiation using ultrasonic sensors and localization problem using cameras’ images. Sensors’ data in these experiments are fused by fusion rules based on the obtained models and then the fusion outcomes are evaluated by the associated uncertainty measures in uncertainty point of view. The results are discussed to illustrate high performance of the new developed generalized aggregate uncertainty measures in the introduced decision making system

Defense Date: Sunday, October 23, 2011 at 5 PM
Room 303, Electrical Engineering Department