Design of neural network based MRAC. (2018)

• Record Type: Journal Article
• Authors: Sethi, Dixit; Sharma, Swati; Kumar, Jagdish; Khanna, Rintu
• Is Part Of: International journal of system control and information processing. Volume 2:Number 4(2018)
• Page Start: 288
• Page End: 304
• Publication Date: 2018
• Subjects: MRAC -- model reference adaptive controller -- AMRAC -- advanced-MRAC -- NN-MRAC -- neural network based MRAC -- MIT rule -- Lyapunov Rule -- PID -- MBPNN -- multilayer backpropagation neural network; System design -- Data processing -- Periodicals; Information technology -- Periodicals; 003.5
• Languages: English
• ISSNs: 1759-9334

Abstract

This paper presents the comparison of conventional-MRAC (model reference adaptive controller), Advanced-MRAC and neural network based MRAC (NN-MRAC) scheme. The MIT Rule and Lyapunov Rule is used for the design of controller parameter adaptation laws. The main focus of research is on how to adapt the control actions more effectively to solve the problem of disturbances and non-linearities. Conventional-MRAC alone is unable to handle nonlinearities and disturbances of the plant and to provide a stable and controlled output, we augment an NN controller, in parallel with MRAC controller, to compensate the nonlinearities and disturbances present in the plant and this scheme is called as NN-MRAC scheme. All methods are applied with analytical detail to a chosen single-input/singleoutput (SISO) second order inherently unstable system named Inverted Pendulum with the application of some uncertainties and disturbances. It is clearly seen from the computer simulation results that NN-MRAC system improves the performance of the system effectively.