On The Problem of Cyber-Physical Systems Architectures Synthesis Using Generative Models

Abstract

The paper dwells on the problem of holistic perception of heterogeneous information by a computing system - text, images, formulas, etc. that remains open. It is to the problem of holistic perception that the task of synthesizing the CPS architecture for implementing adaptive control mechanisms can be attributed. This paper discusses experiments on the synthesis of CPS architectures using open pre-trained generative models, on retraining of a neural network model Stable Diffusion, as well as a new model based on the generative-adversarial approach for the synthesis of CPS architectures.