Agent Systems as Modern methods in the Simulation Tools

Abstract

Modern computer simulation tools are based on different mathematical models. The simplest of these models are rank as rigid computational methods like statistics or probability mathematics. To flexible, more complex computational methods we can classify artificial neural systems, genetic algorithm or fuzzy sets. Without doubt usage of this second group of models are becoming more universal. They are classified as methods of artificial intelligence. With regard to their mechanisms of teaching, neural systems are used for the purpose of optimization. They allow to describe different non-linear structure of the data and to classify them in the appropriate way. Statistic methods do not show optimum results in the non-linear spaces and spaces that are dimensionally complex. Agent methods that are used in the intelligent systems make possible to precisely simulate reality. Reality with huge amount of objects that are on the different level of abstraction. They are used for example in rout planning by means of GPS system, in the planning different logistic processes, in the economics, in the forecasting of meteorological phenomenons etc. Common feature of all forecasting methods are mistakes connected with discrepancy of simulation results and real value. In the literature there is no researches in mentioned discipline. Results of optimization presented in the article are obtained using agent systems. These results refer to finding the shortest way to the defined extreme by studying level of route complexity (multiple, amount of extremes), time of studying, mistakes of defining appropriate extreme. To authenticate results in case of different tools to simulation, researches have been conducted with five different simulation computer programs.