Abstract

The aim of the work was to design, implement, and use, in a number of experiments, an abstract software environment (an artificial world)suitable for modelling systems consisting of many moving and interactingobjects distributed in space. The environment, named DigiHive, is directed towards modeling of complex systems manifested by processes of self-organization, self-reproduction and self modifications. The environment is mainly aimed at modelling and discovering the basic general propertiesof complex systems consisting of many elements and less at modelling of specific phenomena of real systems.The environment is a two-dimensional space in which particles and complexes of particles exist, move, and collide. The collisions may result in creating or removing bonds between particles. On a higher level of interactions, complexes of particles may selectively modify the structures of other particles and complexes of particles according to functions encoded in them. The latter property opens unlimited possibilities of modelling the behavior of complex systems - systems which elements can mutually modify each other changing their structures and this way change their functions. This in turn can lead to the emergence of new functions and interactions which is a crucially novel property of the designed environment. The properties of the environment constitute an artificial world model where many basic ideas of biological processes, living organisms or software agents' interactions may be investigated by computer simulations. The system allows to simulate the evolution of various structures, especially self-reproducing ''organisms'' but without the initially defined fitness criteria (which is the basic need in standard evolutionary algorithms). The criteria are implicitly defined in the rules (''physics'') of the environment. The functions embedded in the structures of complexes are expressed in a specially designed Prolog-like language. The key feature of the language is the property that small changes in the code of a program should usually lead to relatively small changes in the program's behaviour. Such a property of the language is crucial while using the system to simulate the evolutionary, spontaneous development of complex structures. Using the DigiHive environment, two main groups of experiments were designed and performed: 1. Various manifestations of emergent phenomena - self-organization of particles starting from initial state.2. A universal constructor - being the first step to simulate various self-reproduction strategy and its dynamics.

Author (1)