Distributed Environments for Evolutionary Algorithms by means of Multi-Agent Applications

Advanced modeling of control and optimization in management science often leads to a computational complexity which cannot be handled by traditional algorithms and computer systems. On this background the paper develops a general approach to combine the power of distribution and parallelism in natural systems and modern distributed and parallel computing systems. The link between these topics is reached by the concept of multi-agent systems. We show how to build a genetic agent system upon a base model of distribution. Computational performance of this system is presented by a sample application to production scheduling and a runtime analysis of distributed and parallel processing.