Composite adaptive control with fast convergence for multilayer neural network

Abstract

Summary A composite adaptive control (CAC) that combines the benefits of direct and indirect adaptive controls has better parameter adaptation and control response. Multilayer neural networks (NNs) can be employed to enhance a model's representation capacity, but previous composite adaptive approaches cannot easily train the model due to its nonlinearities. A novel CAC is therefore developed in this study to tackle the above limitations. A modified robust version is adopted by focusing on the direct adaptive part to enhance robustness of adaption. Then, the indirect parameter adaptive law is improved by adopting a small learning rate in which a multistep adaption update is executed in one control interval. Moreover, multistep prediction errors are implemented to guarantee the consistency of the approximation errors, and an experience replay technique is adopted to attenuate the requirement of persistent excitation conditions. These improvements not only accelerate the convergence process but also smoothen the updating of NN parameters. Given that a nonlinear plant with MIMO strict‐feedback structure is considered, the proposed CAC is integrated into the backstepping framework. The uniformly bounded property of the tracking errors and the approximation errors is proven by Lyapunov theory. The superiority of the proposed method and the roles of these improvements are demonstrated by comparative simulations.

Neural networks-based command filtering control for a table-mount experimental helicopter

“…For example, image processing, signal processing, pattern recognition, aircraft control and robot control. For nonlinear uncertain function existing in nonlinear systems [19][20][21][22] , the approximation of neural networks can be utilized to estimate this nonlinear uncertain function. In order to improve the performance of the neural network and further control one or more nonlinear systems more accurately, the researchers optimized the parameters of the neural network in [19,21] .…”

“…For nonlinear uncertain function existing in nonlinear systems [19][20][21][22] , the approximation of neural networks can be utilized to estimate this nonlinear uncertain function. In order to improve the performance of the neural network and further control one or more nonlinear systems more accurately, the researchers optimized the parameters of the neural network in [19,21] . Additionally, many researchers use neural networks to estimate the uncertain items in the 3-DOF helicopter system model in order to improve the performance of the control algorithm [17,23] .…”

Multilayer neurocontrol of high‐order uncertain nonlinear systems with active disturbance rejection

“…After substituting (25) into (24) and meanwhile adding and subtracting the terms W T l φl + ŴT l φl , we have…”

“…Moreover, a MNN adaptive robust controller has been proposed 23 . In addition, a MNN composite adaptive controller has been proposed for a multiple‐input multiple‐output system 24 . However, exogenous disturbances have not been considered in Reference 24, making it conservative for practical applications.…”

Fast and stable composite learning via high‐order optimization

“…The direct adaption, which mainly applies tracking errors as a feedback, enables the stability and robustness of the closed‐loop system, whereas the indirect adaption law estimates parameters by minimizing the prediction errors of identification models 17 . The superiority of CLC in ensuring performance improvement has been demonstrated in many control designs, such as the latest results of References 18‐23. Likewise, fast and high‐precision composite adaption with large learning rates causes unsmooth parameter updates, chattering control responses or even deteriorate control performance.…”