Interaction of cognitive and reactive agents in an intelligent computing system: principles of organization

Purpose of research. The problem statement is inspired, among other things, by the fact of the evolution the existing World Wide Web into the Semantic Web. The aim of this work is to develop a multi-agent environment with agents of different types: cognitive and deductive reactive. It is shown that the knowledge and beliefs storage of cognitive BDI agents (BDI is an abbreviation for Belief–Desire–Intention) in the form of facts of an extensional database can be used to obtain new knowledge and beliefs by deductive inference in the environment of an intensional database of agents possessing deductive presumptions. Methods. The operation of an intelligent system is described by an incompletely defined conceptual graph and a system of production rules suitable for representing deductive presumptions of agents in a high-level declarativeimperative programming language. Reactive agents implement their rational behaviors based on deductive abilities, which are understood as the ability to build correct conclusions. Results. The implementation of an intelligent agent-based system is proposed. At the conceptual level, the architecture of the intelligent agent-based system is proposed to be represented by three sublevels. The cognition of agents working at the first and second sublevels is associated with modalities. Based on the formulated interpretations at the third sublevel, reactive agents implement their rational behavior based on deductive abilities, which imply the ability to build correct conclusions. Conclusion . It is shown that the software implementation of cognitive presumptions can be performed using conventional imperative programming languages and, possibly, data manipulation languages. The implementation of the functions of deductive reactive agents when performing operations with conceptual graphs is demonstrated using the example of filling in missing relations in a specific subject area.