Wavelet Transform-Based Frequency Tuning ILC

Zhang, Bin; Wang, Dongliang; Ye, Yongqiang

doi:10.1109/tsmcb.2004.841411

Cited by 39 publications

(26 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wavelet theory is an analytical method used in many scientific fields [10][11][12][13][14][15][16][17][18]. A. Raheja introduced a wavelet based multiresolution algorithm by extending the concept of switching resolutions in both image and data spaces [10].…”

Section: Identify Noise Variance Using Waveletmentioning

confidence: 99%

“…Eung proposed a new method for designing FIR triplet wavelet filter banks, while a cutoff frequency needs to be defined manually [13]. Zhang introduced a cutoff frequency tuning method by using discrete wavelet packet decomposition [14]. In [15], adaptive filters are designed with specific cutoff frequencies for each sub-signal which is decomposed from profile signal by using stationary wavelet transform (SWT) algorithm.…”

Section: Identify Noise Variance Using Waveletmentioning

confidence: 99%

See 1 more Smart Citation

Automatic determination of cutoff frequency for filter design using neuro-fuzzy systems

Wang

Chen

2015

Neurocomputing

View full text Add to dashboard Cite

Section: Identify Noise Variance Using Waveletmentioning

confidence: 99%

Section: Identify Noise Variance Using Waveletmentioning

confidence: 99%

Automatic determination of cutoff frequency for filter design using neuro-fuzzy systems

Wang

Chen

2015

Neurocomputing

View full text Add to dashboard Cite

“…Here, 3.5 Hz is the learnable bandwidth of the conventional ILC system [29]. Figure 6 shows the root mean-square (RMS) error.…”

Section: Trajectory With Zero Initial Position Offsetmentioning

confidence: 99%

Pseudo‐downsampled iterative learning control

Zhang

Wang

Zhou

et al. 2007

Intl J Robust & Nonlinear

View full text Add to dashboard Cite

SUMMARYIn this paper, a simple and effective multirate iterative learning control (ILC), referred as pseudodownsampled ILC, is proposed to deal with initial state error. This scheme downsamples the tracking error and input signals collected from the feedback control system before they are used in the ILC learning law. The output of the ILC is interpolated to generate the input for the next cycle. Analysis shows that the exponential decay of the tracking error can be expected and convergence condition can be ensured by downsampling. Other advantages of the proposed pseudo-downsampled ILC include no need for a filter design and reduction of memory size and computation. Experimental results demonstrate the effectiveness of the proposed scheme.

show abstract

“…These shortcomings can be improved using time-frequency analysis of the feedforward signal [28,41]. The high energy of the error signal is concentrated very locally in time, the learning process outside these intervals leads to an amplification of the present noise.…”

mentioning

confidence: 99%

“…A timevarying robustness filter can adapt to the momentary frequency content of the feedforward signal and allow high frequency system dynamics at the appropriate time instants. A time-varying Q filter can be designed using a time-frequency analysis [13,42,41]. The time-frequency analysis of [13] is performed using the Wigner-Ville decomposition (WVD).…”

mentioning

confidence: 99%

Iterative learning control with wavelet filtering

Merry

Molengraft

Steinbuch

2007

Intl J Robust & Nonlinear

View full text Add to dashboard Cite

SummaryIterative learning control (ILC) has proven to be a powerful method to derive a high performance feedforward signal for systems that perform repetitive tasks. The feedforward signal is derived through several iterations, based on the tracking error of a feedback controlled system. In practical applications, the tracking error consists of a repetitive part and a non-repetitive part. ILC only compensates for the repetitive part of the error. The non-repetitive part also enters the learning scheme and affects the tracking performance. The problem definition of this thesis is defined as:Determine how non-repetitive errors affect the tracking error and the learned feedforward signal of ILC and design a method to eliminate the effect of non-repetitive errors.An expression for the tracking error of an arbitrary iteration is derived as a function of the reference, load disturbances and measurement disturbances. With this expression, the influence of disturbances on ILC can be analyzed. The expression also shows the influence of model uncertainties on the tracking error. The disturbances of the last two iterations appear to have the greatest influence on the tracking error, earlier disturbances only have a limited affect on the tracking error.In order to reduce the effect of disturbances, three filtering methods are developed. The filtering methods are based on the discrete wavelet transform (DWT). The DWT decomposes a signal into wavelet coefficients in various frequency bands. From these wavelet coefficients, the original signal can be reconstructed again. The DWT performs a local time-frequency decomposition. The nonrepetitive part of the disturbances can be removed by changing the wavelet coefficients, i.e. the frequency content of the signal at specific time instants. The designed wavelet filtering methods calculate a measure for the similarity between two sets of wavelet coefficients at an equal iteration of ILC. The repetitive part of the error results in equal wavelet coefficients, the non-repetitive part in different wavelet coefficients. The repetitive part is identified by adjusting the wavelet coefficients based on thresholding of the similarity measure. The adjusted wavelet coefficients are used to reconstruct a disturbance free error signal, which is the input for ILC.The wavelet filtering methods differ in the amount of simulations/experiments (runs) required each iteration of ILC and the way the error signals are obtained. The first method uses two runs each iteration. The two error signals are decomposed into two sets of wavelet coefficients which are used to calculated the similarity measure. The second method is recursive and performs only one run each iteration. The second error signal is constructed using the error of the previous iteration and the feedforward update signal of the present iteration, which is linear filtered with a model of the process sensitivity. For the third method, four runs are performed each iteration, which makes it possible to calculate two similarity criteria and to b...

show abstract

Wavelet Transform-Based Frequency Tuning ILC

Cited by 39 publications

References 17 publications

Automatic determination of cutoff frequency for filter design using neuro-fuzzy systems

Automatic determination of cutoff frequency for filter design using neuro-fuzzy systems

Pseudo‐downsampled iterative learning control

Iterative learning control with wavelet filtering

Contact Info

Product

Resources

About