Wavelet-based Disturbance Classification with Robot Ann Application Example

Abstract

In a certain robot control benchmark it is known that certain types of disturbances occur. This paper presents a classifier that gives a fuzzy estimate of the disturbance that is currently affecting the process. The classifier uses the discrete wavelet transform to extract features from the measurement signal. Based on the features, a fuzzy inference system gives an estimate of the proportion of different disturbances that are present in the signal. The output of the classifier is used to select the controller such that a controller that is tuned for a particular disturbance is used for controlling the process during the disturbance