Cognitive regulators: soft computing technologies and the information-thermodynamic law of intelligent control self-organization

The paper considers a methodology for designing intelligent cognitive control systems for complex dynamic systems. There are brief descriptions of informational and thermodynamic approaches that unite dynamic stability, controllability and robustness criteria under a homogeneous condition. The authors indicate the problems of training and adaptation of a fuzzy controller, which are relevant in modern control theory. Many existing solutions use artificial neural network models based on the backpropagation algorithm (BP), the Cohen multilayer structure, etc. Unfortunately, such algorithms do not guarantee the required level of reliability and control accuracy in complex unforeseen situations. These schemes work successfully if the control task is performed in the absence of underdetermined stochastic noise in the envi-ronment, in sensors, in the control loop, etc. The paper proposes one of the solutions to the problem of developing a cognitive control system, which proposes a constructive solution to the problems of designing knowledge bases and intelligent robust cognitive control in a given problem-oriented application. There is a comparison of various types of regulators, including an intelligent regulator based on emotional brain training. The paper describes the advantages of designing robust knowledge bases based on the software-algorithmic complex Soft Computing Optimizer based on fuzzy logic. The paper also considers one of the key tasks of modern robotics that is the development of technologies for cognitive mechanical interaction, which makes it possible to implement intelligent control functions through the redistribution of knowledge and control at the program level. A practical example shows the effectiveness of the proposed hybrid cognitive control system, which increases the accuracy and reliability of recognizing mental commands.