Abstract : The article provides an overview of the optimal control prediction framework for non-linear CPS and DCS in automation. The core issue is the fact that the present architecture of modern automation systems does not behave like a closed loop control system anymore. The design of these systems combines sensors, actuators, edge controllers, communication channels, digital twin, software-as-a-service, human involvement, and states of cyber-security. Thus, control design must not only address accuracy but should be a multicriteria design problem that accounts for reliability, uncertainty, probability mass, delay in communication, and energy costs. Four levels are considered in the proposed framework. The first level refers to forecasting based on hybrid ARIMA-ML model for a short horizon. The second level is concerned with estimating of risk states using Markov model/HMM. The contribution is a simulation-ready mathematical architecture in which each node solves a local Hamiltonian problem using predicted states, neighbour information and reliability constraints, while the global CPS behavior emerges through networked local decisions. The paper formulates the nonlinear dynamics, cost functional, Hamiltonian conditions, reliability constraints and evaluation protocol for smart factories, smart grids, intelligent buildings and smart campuses. The framework is positioned as a bridge between predictive maintenance and optimal distributed automation control.
Vasil’ev et al. (Wed,) studied this question.