FORECASTING OF THE DISTRIBUTED CYBER-PHYSICAL SYSTEMS STATE

Abstract

The important component for the solution of the problem of decentralized coordination of the distributed cyber-physical systems (DCPS) control is obtaining of the primary information, needed for searching the optimal control. Such information may be divided into conditional constant – the parameters of the controlled object, and variable – state of the controlled object. Determination of the object parameters is carried out by means of identification problem solution. Object persistence and correspondingly, the possibility of changing its elements state during the coordination cycle stipulates the need to forecast the processes in DCPS. In greater part of the research, dealing with the forecasting problem methods are considered where in this or that way expert assessments and conclusions are used. This concerns mainly social-economic processes, forecasts in the sphere of medicine, education, etc.

In the given study the forecasts for the cyber-physical systems are considered, although, in this case models of physical processes and formal methods of the forecasting  play far more important role , however, at certain stages expert assessments are also used, in particular, regarding the ranges of possible change of the parameters, list of influencing factors, etc. The prediction method of the state of the distributed cyber-physical systems with the continuous objects on the base of space-time spectral approach to the prediction of DCPS state with continuous and discrete object states has been improved, study of the characteristic of DCPS state prediction has been performed. The possibility of the realization of the prediction method, using the machine learning and simulation modelling has been considered. The expediency of the prediction depth as with the increasing of the depth (interval) of the prediction the fuzziness of the prediction results increases. At the same time, computational resources and time are spent for the prediction. Gradually the situation arises, when the positive effect of the prediction becomes less than the expenses for its realization, this determines the expedient maximum depth of the prediction.