Adaptive nonlinear control using input normalized neural networks

Leeghim, Henzeh; Seo, In-Ho; Bang, Hyochoong

doi:10.1007/s12206-007-1119-1

Cited by 18 publications

(7 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In contrast to the neuro-predictive method, ANN feedback linearization has been widely used to control second-order mechanical systems [87,88,90,92], and first-order processes/mechanical systems [86,91,94] have also been controlled by this method.…”

Section: Ann Feedback Linearization Control Systemsmentioning

confidence: 99%

“…for uncertainty compensation 23, 89; these cases are not considered as ANN feedback linearization systems in this paper. MLPs 86–88, 90, 91 and RBFNs 3, 92, both in the recurrent form, are the most popular neural networks in ANN feedback linearization. In many cases, using the capability of ANNs in on‐line learning, the designed ANN feedback linearization control system is adaptive 3, 90, 92–94.…”

Section: An Introduction To Neuro Controlmentioning

confidence: 99%

“…MLPs 86–88, 90, 91 and RBFNs 3, 92, both in the recurrent form, are the most popular neural networks in ANN feedback linearization. In many cases, using the capability of ANNs in on‐line learning, the designed ANN feedback linearization control system is adaptive 3, 90, 92–94. In these control systems, the ANN's structure and learning laws can be defined so that the control system is stable 3, 90–92.…”

Section: An Introduction To Neuro Controlmentioning

confidence: 99%

“…In many cases, using the capability of ANNs in on‐line learning, the designed ANN feedback linearization control system is adaptive 3, 90, 92–94. In these control systems, the ANN's structure and learning laws can be defined so that the control system is stable 3, 90–92. In contrast to the neuro‐predictive method, ANN feedback linearization has been widely used to control second‐order mechanical systems 87, 88, 90, 92, and first‐order processes/mechanical systems 86, 91, 94 have also been controlled by this method.…”

Section: An Introduction To Neuro Controlmentioning

confidence: 99%

See 3 more Smart Citations

A critical review of the most popular types of neuro control

Mohammadzaheri

Chen

Grainger

2011

Asian Journal of Control

View full text Add to dashboard Cite

In this review article, the most popular types of neural network control systems are briefly introduced and their main features are reviewed. Neuro control systems are defined as control systems in which at least one artificial neural network (ANN) is directly involved in generating the control command. Initially, neural networks were mostly used to model system dynamics inversely to produce a control command which pushes the system towards a desired or reference value of the output (1989). At the next stage, neural networks were trained to track a reference model, and ANN model reference control appeared (1990). In that method, ANNs were used to extend the application of adaptive reference model control, which was a well-known control technique. This attitude towards the extension of the application of well-known control methods using ANNs was followed by the development of ANN model-predictive (1991), ANN sliding mode (1994) and ANN feedback linearization (1995) techniques. As the first category of neuro controllers, inverse dynamics ANN controllers were frequently used to form a control system together with other controllers, but this attitude faded as other types of ANN control systems were developed. However, recently, this approach has been revived. In the last decade, control system designers started to use ANNs to compensate/cancel undesired or uncertain parts of systems' dynamics to facilitate the use of well-known conventional control systems. The resultant control system usually includes two or three controllers. In this paper, applications of different ANN control systems are also addressed.Key Words: Control, neural network, adaptive, feedback linearization, predictive, model reference, perceptron, radial basis. I. ARTIFICIAL NEURAL NETWORKS USABLE IN CONTROLArtificial neural networks (ANNs) are special mathematical models inspired by human neural networks. An ANN usually includes neurons, connections and biases. Neurons are arranged in 'layers'. There are a variety of neural networks suitable for different purposes [1,2]. In neuro control, it is a difficult and unsolved problem to find the best ANN structure for each specific application; thus, a fairly large ANN is usually employed to deal with relatively complex approximation problems [3]. In this paper, the most common types of artificial neural networks in the area of control are introduced: multi layer perceptrons (MLPs) and radial basis function networks (RBFNs). Both of these are known as universal approximators of systems [4,5]. The advantage of RBFNs is their quick training process compared with MLPs; however, for complicated systems with many inputs, the number of neurons in RBFNs are usually considerably higher than MLPs [1,2]. Regardless of the main structure, if the output of an ANN depends not only on the current q

show abstract

Section: Ann Feedback Linearization Control Systemsmentioning

confidence: 99%

Section: An Introduction To Neuro Controlmentioning

confidence: 99%

Section: An Introduction To Neuro Controlmentioning

confidence: 99%

Section: An Introduction To Neuro Controlmentioning

confidence: 99%

See 2 more Smart Citations

A critical review of the most popular types of neuro control

Mohammadzaheri

Chen

Grainger

2011

Asian Journal of Control

View full text Add to dashboard Cite

show abstract

“…The input and the output data used for training the NN are normalized. This not only makes various observations equally important but also helps reduce the training error and the training time [13,14]. A general three-layered NN architecture is shown in Fig.…”

Section: B Neural Networkmentioning

confidence: 99%

Neural Partial Differential Method for Extracting Aerodynamic Derivatives from Flight Data

Das

Kuttieri

Sinha

et al. 2010

Journal of Guidance, Control, and Dynamics

View full text Add to dashboard Cite

Although the classical approach based on output or equation error methods is commonly applied to estimate stability and control derivatives from flight data, an alternative approach based on a neural network with universal nonlinear mapping capability is preferred in some applications. The first approach yields explicit models in terms of physically interpretable aerodynamic derivatives, whereas the other leads to implicit models in general. The two approaches being characteristically different, the comparison of the results in terms of aerodynamic derivatives pose difficulties. An approach based on finite differences (termed as the delta method), applied to a trained neural network, has been reported in the literature to overcome the difficulties. In this paper, yet another approach based on the partial differential of neural output is suggested and demonstrated. The performance of the two approaches (namely, the finite difference and the partial differential approach) is evaluated in terms of the estimated aerodynamic derivatives. It is shown that the two methods give comparable results; however, the partial differential approach provides first-order aerodynamic derivatives more accurately. Moreover, the results of the partial differential approach are comparable with the results obtained from the equation error method. Moreover, gradient descent and scaled conjugate gradient methods are evaluated in terms of the mean square error for training the neural network. The proposed partial differential approach does not suffer from numerical instability and provides higher-order derivatives; therefore, it can be extended to the nonlinear aerodynamic regime and to unstable aircraft. Nomenclature C l = nondimensional rolling moment coefficient C n = nondimensional yawing moment coefficient C Y = nondimensional side force coefficient C Y 0 , C l 0 , C n 0 = zero aerodynamic term in the linear aerodynamic model d i = normalized observed or measured value of ith output E:= error function E 0 := gradient of the error function f: = activation function g:= activation function L = rolling moment, Nm N = yawing moment, Nm p = roll rate, rad=s p = normalized roll rate q = pitch rate, rad=s r = yaw rate, rad=s r = normalized yaw rateŨ = weight matrix connecting the input layer and the first hidden layer V = weight matrix connecting the first hidden layer and the second hidden layer W = weight matrix connecting the second hidden layer and the output layer w = weight vector containing all weights of the neural network x = output vector of the first hidden layer Y = side force, Ñ y = output vector of the second hidden layer z i;norm max = upper limit of normalization for the ith output z i;max = maximum value of the ith output z i;min = minimum value of the ith output z i;norm min = lower limit of normalization for the ith output z = output vector of the neural network = sideslip, rad a = aileron deflection, rad r = rudder deflection, rad i = ith element of the inputs to the second hidden layer neuroñ = input vector to the neural n...

show abstract

A Comparative Approach for Email Spam Detection Using Deep Learning

Singh

Chatterjee

2021

Algorithms for Intelligent Systems

View full text Add to dashboard Cite

Adaptive nonlinear control using input normalized neural networks

Cited by 18 publications

References 24 publications

A critical review of the most popular types of neuro control

A critical review of the most popular types of neuro control

Neural Partial Differential Method for Extracting Aerodynamic Derivatives from Flight Data

A Comparative Approach for Email Spam Detection Using Deep Learning

Contact Info

Product

Resources

About