EVOALG: Grundlagen und Anwendungen evolutionaerer Algorithmen Schlussbericht

At the beginning of the project only few theoretical results regarding evolutionary algorithms (EAs) were known, so that a goal-oriented development of EAs for specific problems could not rest on a theoretical basis. The goal of this project was the development of a theoretical foundation of EAs and their application to real world problems in industry. The theoretical results should be transferred to practical use, validated and perhaps improved with the help of the industrial partner. The theoretical foundation rests on the theory of stochastic processes. The real world problems of the industrial partner led to the development parallel EAs which were realized via the parallel interface tool PVM. The theoretical foundation led to parallel special purpose EAs tailored to the optimization of optical multi-layer coatings, nuclear core loading plans, and routing tables of private nets in telecommunication. Solutions found by human experts or previously used software were consistently improved. Moreover, we developed a preliminary graphics-based software tool for interactive evolution which is intended to support design tasks. This tool attracted much attention, even in the broad public. The theoretical foundation was the basis for the goal-oriented development of EAs for industrial problems. Moreover, the coupling of optimization and simulation for solving industrial problems turned out to be extremely fruitful. Because of the inherent parallelism of EAs, parallel MIMD computers as well as workstation clusters could be used in a profitable manner. This approach should prove successful in similar applications and it will gain practical relevance in future. (orig.)SIGLEAvailable from TIB Hannover: F98B1261+a / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekBundesministerium fuer Bildung, Wissenschaft, Forschung und Technologie, Bonn (Germany)DEGerman