NSF Award
9409358
9409358 Li This proposal deals with an important class of adaptive schemes for state/parameter estimation, known as multiple model (MM) estimation. These schemes are characterized by using multiple models and are particularly good for problems involving both structural as well as parametrical changes. Applications of MM estimation can be found in many areas, such as failure detection/isolation, maneuvering target tracking, air traffic control, and adaptive filtering. An intelligent framework of the following three variable structure schemes will be developed in particular: active model set, model-set switching, and adaptive grid. Theoretical work on model set selection will also be carried out, which will provide a solid basis and a useful guideline for design of practical MM estimators. Moreover, a theoretical framework for the MM estimation will be established based on graph theory. This will lead to systematic treatment of model set evolution (adaptation) as well as better understanding of the existing schemes. The developed schemes will be applied to practical problems and their performance and computation will be compared with those of the existing schemes. ***
