Evolutionary Method of Constructing Artificial Intelligence Systems
- PDF / 175,260 Bytes
- 9 Pages / 594 x 792 pts Page_size
- 31 Downloads / 192 Views
CYBERNETICS EVOLUTIONARY METHOD OF CONSTRUCTING ARTIFICIAL INTELLIGENCE SYSTEMS A. V. Anisimov,1† O. O. Marchenko,1‡ and V. R. Zemlianskyi1††
UDC 681.3
Abstract. An evolutionary model of constructing artificial intelligence is presented, which is destined for designing and developing intelligent systems. The model allows describing a variety of subject areas with constructing knowledge bases. It has universal means to formally describe tasks and environments for implementing computational processes to solve them. The key basic element of the proposed model is the so-called ALF, i.e., an intelligent agent with the abilities to self-learning, communication, self-organization, and joint actions with similar agents. The development of ALF agents is based on evolutionary principles implemented using genetic algorithms. The proposed approach is implemented in the form of a game model. The developed structure and functionality of ALF agents stipulate the flexibility and efficiency of the model, which is confirmed by experiments. Keywords: artificial intelligence, multiagent system, evolutionary programming. INTRODUCTION The developed model is destined for solving a number of theoretical and practical problems of artificial intelligence. Proceeding from this, the formulation of a specification for constructing the model includes the following requirements imposed on it: · the presence of tools for the formal description of knowledge that represents the outside world or some subject area and is comparable with ontologies by expressiveness and universality [1]; · the possibility of formal description of a problem, i.e., its statement; · the presence of an environment for efficiently searching for a problem solution. The listed properties, especially the two latter, are implemented in the programming language Prolog [2]. The proposed model is based on principles of evolutionary systems, which makes it flexible, universal, and adaptive. The basic element of the model is the so-called Artificial Life Form (ALF), i.e., an intelligent autonomous agent with the ability to learn in the process of search for a problem solution (i.e., searching for the objective), to adapt to a variable computing environment, and also to plan and replan its actions and jointly act with other agents by communicating with them. In this sense, the model being considered is similar to the multiagent approach but, to achieve its universality, the emphasis is placed on the formation of a special ALF agent objective function to which all its possible actions, plans, and problems are subordinated. The objective function of an ALF agent is the coefficient of its “survivability,” i.e., its main objective is to stay “alive” in its environment as long as possible. In what follows, the problem statement is formulated and thus the corresponding configuration of the computing environment of the model is defined and the process of computation of 1
Taras Shevchenko National University of Kyiv, Kyiv, Ukraine, †[email protected]; ‡[email protected]; slava.zem
Data Loading...
