Multilevel, hybrid intellignet control system

Abstract

Intelligent controls have been noticed in control theory because of their powerful control capability. However, to make an intelligent control system perform some of the complex and laborious tasks currently done by people, those methodologies of intelligent controls should be integrated into an intelligent control system with hierarchy and modularity in a systematic manner. This thesis provides a frame-work to achieve this goal based on several control concepts - an event-based intelligent control, a fuzzy logic control, a PID control, and two methodologiesneural networks, Genetic Algorithms (GAs). We employed a multilevel control structure to hierarchically manage environmental informations with three levels such as continuous one, discrete one, and fused one. Also, a hybrid control methodology was taken to control any plants composed of discrete and continuous modules. An event-based intelligent controller which is the main framework of our approach plays a high level supervising controller. That is, this control paradigm takes discrete environmental informations and makes decision using an endomorphic plant model and a current scheduled control task. The endomorphic plant model should provide our control system with an adaptability about environmental variation. An endomorphic neural network model is employed to give this adaptability. The results of decision are sent to low-level control modules to track predetermined control sequences. Control tasks, i.e., control objectives are managed by a higher-level modules, a scheduler and a planner. Several new methodologies related with the lower-level modules are devised and applied to our system to increase the performances of the modules and provide our system with high autonomy. We introduced an adaptable PID-plus Bang-Bang control with a predictive model using a neural network and with a gradient based prediction model; a fuzzy logic control method with multi-shaped output fuzzy sets, its defuzzification meth...