The contributions of this thesis address both researchers in field of model and controller reduction and they are essentially focused on the problem of instability in both fields.
The first contribution concerns model reduction methods based on similarity transformations. For a given model a general framework is proposed that parameterizes a large set of reduced models that preserve the stability of the original model. As an application of this result, it is shown how different model reduction methods can be modified, if they fail to maintain stability. The basic idea in this part is also exploited as a guideline to deal with the main objective of this dissertation which is preserving stability of the closed loop system in case of controller reduction.
The second contribution of this thesis is isolating the problem of guaranteeing the internal stability of the reduced closed loop system from the problem of transfer matrix matching. Our treatment of stability is based on generalized Gramians. The approach here, utilizes the theoretic framework based on strict Lyapunov inequalities. Then for a given controller a framework is proposed that parameterizes a set of reduced controllers that preserve the stability of the closed loop system. In addition, a sufficient condition for the existence of such a framework is derived. An interesting application of these results is an algorithm to preserve stability in currently available controller reduction approaches that suffer from the lack of stability, by slightly modifying the corresponding similarity transformations.
In addition in an individual chapter, the effect of controller reduction on embedded systems is deeply investigated. This chapter gives an insight into the effectiveness of controller reduction with regard to the size and processing speed of embedded controllers after reduction.
«
The contributions of this thesis address both researchers in field of model and controller reduction and they are essentially focused on the problem of instability in both fields.
The first contribution concerns model reduction methods based on similarity transformations. For a given model a general framework is proposed that parameterizes a large set of reduced models that preserve the stability of the original model. As an application of this result, it is shown how different model reduction...
»