Identification and Control of Networked Control Systems (NCSs)

Abstract

Networked Control Systems (NCSs) are the systems wherein the control loops are closed using real-time networks. NCSs are cheaper, flexible and reliable compared to traditional wired control systems. Thus NCSs are used for a wide variety of applications including manufacturing plants, unmanned vehicles and intelligent traffic systems. However, the presence of networks in NCSs cause network-induced delays and packet loss which in turn degrade the performance of the system. In this thesis, methods for delay and packet loss compensations are proposed and compared. It is found that random delay compensation and packet loss compensation using cubic spline interpolation are the more accurate methods. These methods are implemented for system identification using the Modified Least Squares (MLS) and Orthogonal Least Squares (OLS) algorithms. The MLS algorithm is applied to a simple linear system to compare the accuracy of the compensation methods whereas the OLS algorithm is applied to two linear and three non linear NCS examples for several scenarios (with different levels of noise and packet loss). The identification results of all the examples show a similar trend as the same model terms are identified but the values of the coefficients can vary when the packet loss increases. The validation results show that the validation plots are overlapping and the correlation plots are within confidence intervals for all the examples. The identified models are used for design of Model Predictive Controllers (MPCs) to control systems so that the desired outputs are achieved. The results indicate that though the performance of MPCs is affected by packet loss, the NCSs are generally able to attain the desired output for all the scenarios. Therefore, techniques for identification as well as control of linear and non linear Networked Control Systems (NCSs) are developed in this thesis.