C:\moje zbiory\viola\Enformatica_Praga_2006\Praga_2006.dvi

The present paper is oriented to classification and application of agent technique in simulation of anticipatory systems, namely those that use simulation models for the aid of anticipation. The main ideas root in the fact that the best way for description of computer simulation models is the technique of describing the simulated system itself (and the translation into the computer code is provided as automatic), and that the anticipation itself is often nested. Keywords—Agents, Anticipatory systems, Discrete event simulation, Simula, Taxonomy. I. PROBLEMS IN PROGRAMMING OF COMPUTER SIMULATION MODELS AND THEIR SOLUTION OMPUTER Simulation is the best (and may be said the unique) technique for exact studying of the complex dynamic systems, i.e. for the complex systems that are viewed to change their state during a Newtonian (non-relativistic) time flow. For studying dynamic systems that are not complex, simulation is a very expensive (therefore unfitting) technique. Simulation is based on experimenting with the model [1]. In order to reflect the causality relations during the experimenting, the model is demanded to satisfy the following condition: When events E1 and E2 coming in the model, respectively reflect events 1 and 2 coming in the modeled system, and when 1 occurs later than 2, then E2 cannot occur later than E1 in the model. In such a case, E1 usually comes later then E2, but it is possible that both the events come at the same time (it happens namely when the model is rough, but even such models are of use). The reality of simulation is that the complex systems should be mapped to complex computer models with intricate behavior in time, which has to be described as algorithms handling with tangled structure of a lot of data. Moreover, a small change of the simulated system can be frequently discovered during the application of simulation; although such a change could be described by a simple sentence, the corresponding modification of the computer model is often so large that it is comparable with programming a new model from scratch. In order to solve such problems concerning the programming of simulation models, an attempt was accepted and Manuscript received June 27, 2006. The paper was elaborated under support of the grant No. 201/06/0612 of the Grant agency of Czech Republic. Eugene Kindler is with Ostrava University Faculty of Sciences, Czech Republic (phone: +420-221-914-286; fax: +420-221-914-323; e-mail: evzen.kindler@ mff.cuni.cz). elaborated in two ways, which will be shortly described in the next lines; the attempt can be characterized so that instead forcing the author of simulation model M to describe what should happen in the computer C when M exists and operates inside C, the author should describe the simulated system and such a description should be automatically translated into the computer code realizing M. In the present paper, the attempt will be identified as . The last fifty years of computer simulation showed an enormous help of it. The first elaboration of the attempt consists in simulation programming languages (shortly simulation languages). Each of them is oriented to a certain class of dynamic systems that could be well, promptly and without obstacles described is it. During the last 50 years, many tens of simulation languages were designed, correspondingly to many tens of classes of dynamic systems that attracted attention of the simulationists. The translation of the text in a simulation language into the corresponding computer code is a very knotty process, as it should well handle all texts in the simulation language. Implementing it in a form of a compiler is a difficult challenge even for the high level programmers; they are not at disposal, while new classes of systems, which demand their own simulation languages, arise. Already in the sixties of the XX century, this situation stimulated the elaboration of a special technique of representation of concepts, carried by programming language SIMULA [3]-[5]. Many years after, certain ideas existing in SIMULA were followed by the world professional community under title object-oriented programming (shortly OOP) and stepwise reflected in other programming languages like SmallTalk, C++, Eifel, or Java. The ideas were (1) representation of general concepts (commonly under name class) with (2) an absolute freedom to generate any number of their instances to represent individuals carrying the contents of the classes, and with (3) an absolute freedom to specialize such classes to their subclasses to represent concepts with a richer content, (4) the content of the classes consists in their attributes (representing “properties”) and methods (algorithms, representing “abilities”), so that (5) every instance of a class carries all attributes introduced for the class, and is able to perform any method introduced for the class, when it is demanded to do that, and that (6) the specialization of a class consists in adding new attributes and new methods and Agent-Based Simulation of Simulating Anticipatory Systems – Classification