Математическая модель функционирования когнитивной радиосистемы

Abstract

The functioning of a promising cognitive radio system in the context of an electronic conflict is considered. As a result of the analysis, the correspondence of the required intellectual abilities of the cognitive radio system, the tasks to be solved and the intelligent algorithms proposed by the researchers for the realization of these abilities was revealed. The aim of the work is to develop a mathematical model that allows us to describe the adaptation of a conflict–resistant cognitive radio system and assess the complexity of the corresponding algorithms. Two groups of methods were used: general scientific methods – abstraction, generalization, analysis, synthesis, as well as special methods of graph theory, criteria importance theory, algorithm theory and set theory. Two new mathematical models are presented. The first model describes the state of the subscriber of the radio system from a combinatorial point of view, the second describes the functioning of the cognitive radio system in conditions of electronic conflict. The latter model is formalized using graph theory – an r-weighted multigraph is constructed, the vertices of which are identified with the states of the subscribers of the radio system, and the corresponding weight vectors are assigned to the edges. The elements of each vector of weights qualitatively characterize the functioning of the cognitive radio system according to the selected indicators. An example is given with indicators characterizing the functioning of the radio system: stealth, noise immunity, energy efficiency and information transfer rate. The graph model allows you to generalize the description of various methods for obtaining knowledge about the environment and managing radio system resources, and also allows you to evaluate the computational complexity of adaptation algorithms. The graph model also makes it possible to describe the intellectual abilities of a radio system both from the standpoint of a production approach and from the standpoint of reinforcement learning. The given estimates of the computational complexity of some tasks allow us to divide the tasks into two groups: those solved by subscribers and the control subsystem. The consistency of proposals for the separation of computational tasks with the principles of transfer learning is shown.