Nowe ujęcie wybranych zagadnień optymalizacji

Abstract

In solving complex optimization tasks evolutionary algorithms have a leading position. Unusual look at the optimization algorithms presented in the thesis, led to the creation of the new algorithm and work on its development to put its metaphors in a group of artificial life. The resulting algorithms are still the effective optimization algorithms and the proposed approach introduces new properties in their operation. The study presents a new algorithm of observation - as the base algorithm and its metaphors placed in a group of immune algorithm and particle swarm optimization algorithms. Research on the mechanics of these algorithms demonstrated new properties, i.e.: behavior resembling observation, and co-evolution mechanism determines the behavior of independence on influences of the environment. Implementation of the assumptions imposed the need to develop effective mechanism of mutation for immune algorithm. The functions of behavior scenarios were defined for the particle swarm optimization algorithm. A group of immune systems is proposed which is an equivalent to the multi-population system and methods of information exchange between systems in the group are defined. The thesis presents a theoretical background of algorithms’ operation and a simulation study. To check the efficiency of the algorithms the typical test environment for stationary and non-stationary problems were applied. In the study, fractal and multifractal analysis was used and its usefulness was demonstrated in research on behavior of algorithms. Optimization of diagnostic structure of digital circuit is an issue of multimodal optimization and is a particular kind of challenge. A comprehensive approach to test multi-module circuit may lead to new solutions, also in terms of a single module testing. Such concepts are included in this study, basing on an untypical approach to testing multi-module circuit, the conclusion has a strong theoretical base. The original achievements in this dissertation are as follows: a proposal of BIST architecture based on the so-called linear modification, the introduction of the diagnostic structure description, and determination of the theoretical basis of this concept, confirmation of the formulated theoretical basement and simultaneously the verification of the diagnostic efficiency of the proposed solutions by means of simulation methods basing on modeling with using ISCAS’89 benchmark, the demonstration of permanent features of modules during testing, the presentation of a formal description of any diagnostic structure with a description of the optimization framework and the concept of simulation tools used in the current research. Simultaneously, the study shows the original use of a genetic algorithm to give a high efficiency optimization. This part of the study presents a complete system of description of any diagnostic structure with the optimization method. The solutions presented in the dissertation open the way for the further research. This dissertation is composed of two parts, despite of the common basis in a form of evolutionary algorithms, they are present different and closed thematically issues. Keywords: optimization, multi-criteria optimization, multimodal optimization, evolutionary algorithms, genetic algorithms, immune algorithms, particle swarm optimization algorithms, a group of immune system, the algorithm of observation, exchange of genetic material, fractal analysis, multifractal analysis, beset game algorithm, immune algorithm with auto-aggression, stationary problems, non-stationary problems, BIST structure, BIST structure optimization, BIST structure description, multi-modular circuit BIST.