Agent-oriented models, method and tools of compatible problem solvers development for intelligent systems

Abstract

The article is devoted to the development of agent- oriented models, a method and means for developing compatible solvers of problems of intelligent systems capable to solve complex problems. The requirements for such solvers, the model of problem solver that satisfy the requirements, as well as the method and tools for developing and modifying such solvers are considered. The main problem considered in the work is the problem of low consistency of the principles underlying the implementation of various problem solving models. As a consequence, it is difficult to simultaneously use different models for solving problems within the same system when solving the same complex problem, it is practically impossible to reuse the technical solutions implemented in a certain system, in fact, there are no integrated methods and tools for problem solvers development. As a basis for problem solvers design, it is proposed to use the multi-agent approach. The process of any problem solving is proposed to decompose into logical atomic actions, which will ensure compatibility and modifiability of the solvers. The solver is proposed to be considered as a hierarchical system consisting of several interconnected levels, which allows to provide the possibility of independent designing, debugging and verification of components at different levels.