| System identification and adaptive control for nonlinear systems are two difficult problems for control engineering. The reason is that identification of nonlinear systems involves two major steps: the selection of a model structure with certain set of parameters and the selection of an algorithm to estimate these parameters. The later issue usually biases the former one. There are still many unsolved problems related to the implementation and design of adaptive controllers for nonlinear systems. These problems are related to the controller structure, stability, and the adaptation mechanism.; The complexity of modern control system technology and the corresponding associated problems cause difficulties in the analysis and design for such systems using traditional theories. The complexity of the error surfaces for nonlinear systems and the presence of noise are two important factors which increases the chance of failure in building a good model structure for such systems. Intelligent techniques for system identification and adaptive control aim to solve these problems. A new optimization technique borrowed from the field of Artificial Intelligence (AI), Genetic Algorithms (GAs), has yielded an efficient robust behavior while exploring different types of multi-model error surfaces efficiently and finds minimum or maximum points in the search space which are likely to be global in nature.; In this thesis, we explore the usage of GAs to solve the parameter estimation, system identification and adaptive control problems where most traditional approaches might fail to provide appropriate solutions. The problem of estimating nonlinear system parameters is fully addressed. Two methods of building model structures for unknown systems are presented. The first method is developed using a priori knowledge which an engineer can gain from experiments. When this knowledge does not exist, another method is proposed which is quite general and robust in the sense of performance and computations. Also, a new direct adaptive nonlinear controller structure is presented. The controller parameters are adjusted directly based on the output error. This is one of the few available direct adaptive controllers. The reported results demonstrate the motivation behind using GAs to handle system identification and adaptive control engineering problems. |
