Adaptive Neural Network Filter Research Articles

Adaptive fault-tolerant attitude control for hypersonic reentry vehicle subject to complex uncertainties

In this paper, a novel fast attitude adaptive fault-tolerant control (FTC) scheme based on adaptive neural network and command filter is presented for the hypersonic reentry vehicles (HRV) with complex uncertainties which contain parameter uncertainties, un-modeled dynamics, actuator faults, and external disturbances. To improve the performance of closed-loop FTC, command filter and neural network are introduced to reconstruct system nonlinearities that are related to complex uncertainties. Compared with the FTC scheme with only neural network, the FTC scheme with command filter and neural network has fewer controller design parameters so that the computational complexity is decreased and the control efficiency is improved, which is of great significance for HRV. Then, the adaptive backstepping fault-tolerant controller based on command filter and neural network is designed, which can solve the complexity explosion problem in the standard backstepping control and the small uncertainty problem in the backstepping control only containing command filter. Moreover, to improve the approximation accuracy of the neural network-based universal approximator, an adaptive update law of neural network weights is designed by using the convex optimization technique. It is proved that the presented FTC scheme can ensure that the closed-loop control system is stable and the tracking errors are convergent. Finally, simulation results are carried out to verify the superiority and effectiveness of the presented FTC scheme.

Estimation of unknown system states based on an adaptive neural network and Kalman filter

In the field of industry 4.0 the number of sensors increases steadily. The sensor data is often used for system observation and estimation of the system parameters. Typically, Kalman filtering is used for determination of the internal system parameters. Their accuracy and robustness depends on the system knowledge, which is described by differential equations. We propose a self-configurable filter (FNN-EKF) which estimates the internal system behavior without knowledge of the differential equations and the noise power. Our filter is based on Kalman filtering with a constantly adapting neural network for state estimation. Applications are denoising sensor data or time series. Several bouncing ball simulations are realized to compare the estimation performance of the Extended Kalman Filter to the presented FNN-EKF.

Jitter Elimination in Shape Recovery by using Adaptive Neural Network Filter.

Three-dimensional (3D) cameras are expensive because they employ additional charged coupled device sensors and optical elements, e.g., lasers or complicated scanning mirror systems. One passive optical method, shape from focus (SFF), provides an efficient low cost solution for 3D cameras. However, mechanical vibration of the SFF imaging system causes jitter noise along the optical axis, which makes it difficult to obtain accurate shape information of objects. In traditional methods, this error cannot be removed and increases as the estimation of the shape recovery progresses. Therefore, the final 3D shape may be inaccurate. We introduce an accurate depth estimation method using an adaptive neural network (ANN) filter to remove the jitter noise effects. Jitter noise is modeled by both Gaussian distribution and non-Gaussian distribution. Then, focus curves are modeled by quadratic functions. The ANN filter is designed as an optimal estimator restoring the original position of each frame of the input image sequence in the modeled jitter noise, as a pre-processing step before the initial depth map is obtained. The proposed method was evaluated using image sequences of both synthetic and real objects. Experimental results demonstrate that it is reasonably efficient and that its accuracy is comparable with that of existing systems.

An Adaptive Radial Basis Function Neural Network Filter for Noise Reduction in Biomedical Recordings

Electroencephalogram (EEG) recordings often experience interference by different kinds of noise, including white, muscle and baseline, severely limiting its utility. Artificial neural networks (ANNs) are effective and powerful tools for removing interference from EEGs. Several methods have been developed, but ANNs appear to be the most effective for reducing muscle and baseline contamination, especially when the contamination is greater in amplitude than the brain signal. An ANN as a filter for EEG recordings is proposed in this paper, developing a novel framework for investigating and comparing the relative performance of an ANN incorporating real EEG recordings. This method is based on a higher-order statistics-based radial basis function (RBF) network. This ANN improves the results obtained with the conventional EEG filtering techniques: wavelet, singular value decomposition, principal component analysis, adaptive filtering and independent components analysis. Average results for the RBF-based method provided a noise reduction (SIR) of (mean$$\pm $$± SD) $$\mathrm{SIR}=19.3\pm 0.3$$SIR=19.3±0.3 in contrast to traditional compared methods that, for the best case, yielded $$\mathrm{SIR}=15.2\pm 0.3$$SIR=15.2±0.3. The system has been evaluated within a wide range of EEG signals. The present study introduces a new method of reducing all EEG interference signals in one step with low EEG distortion and high noise reduction.

Noise Canceling in Volatility Forecasting Using an Adaptive Neural Network Filter

Abstract Volatility forecasting models are becoming more accurate, but noise looks to be an inseparable part of these forecasts. Nonetheless, using adaptive filters to cancel the noise should help improve the performance of the forecasting models. Adaptive filters have the advantage of changing based on the environment. This feature is vital when they are used along with a model for volatility forecasting and error cancellation in the financial markets. Nonlinear Autoregressive (NAR) neural networks have simple structures, but they are efficient tools in error cancelation systems when working with non-stationary and random walk noise processes. For this research, an adaptive threshold filter is designed to respond to changes in its environment when a GARCH(1,1) model makes errors in its volatility forecast. It is shown that this filter can forecast the noise (errors) in the GARCH(1,1) outputs when there is a non-stationary time series of errors. The model reduces the mean squared errors by 42.9%. A sample portfolio of five stocks from the S&P 500 index from 4/2007 to 12/2010 is studied to illustrate the performance of the model.

Adaptive fuzzy wavelet neural network filter for hand tremor canceling in microsurgery

One of the main problems for effective control of a minimally invasive surgery (MIS) is the imprecision that caused by hand tremor. In this paper, a new kind of nonlinear adaptive filter, the fuzzy wavelet neural network filter (FWNNF), is proposed to alleviate this problem. The proposed FWNNF is inherently a feed forward multilayered connectionist network which can learn by itself according to numerical training data or expert knowledge represented by fuzzy if–then rules. Each rule includes a wavelet function in the consequent part of the rule. In the parameter learning phase, gradient descent method is used to minimize the output error. With the FWNNF, we can model and predict the hand tremor more effectively and improve the precision and reliability in the master–slave robotic system for microsurgery. Extensive simulation results are given and performance comparison with traditional methods is conducted, underlining the effectiveness of the proposed filter and its superior performance over its competing rivals.

Optimization of an adaptive neural network to predict breathing

To determine the optimal configuration and performance of an adaptive feed forward neural network filter to predict breathing in respiratory motion compensation systems for external beam radiation therapy. A two-layer feed forward neural network was trained to predict future breathing amplitudes for 27 recorded breathing histories. The prediction intervals ranged from 100 to 500 ms. The optimal sampling frequency, number of input samples, training rate, and number of training epochs were determined for each breathing history and prediction interval. The overall optimal filter configuration was determined from this parameter survey, and its accuracy for each breathing example was compared to the individually optimal filter setups. Prediction accuracy was also compared to breathing stability as measured by the autocorrelation of the breathing signal. The survey of filter configurations converged on a standard setup for all examples of breathing. For 24 of the 27 breathing histories the accuracy of the standard filter for a 300 ms prediction interval was within a few percent of the individually optimized filter setups; for the remaining three histories the standard filter was 5%-15% less accurate. A standard adaptive neural network filter setup can provide approximately optimal breathing prediction for a wide range of breathing patterns. The filter accuracy has a clear correlation with the stability of breathing.