Sensor classification methods applied to robotics

Abstract

The data obtained from the robotic instrumentation can be redundant due to the multiplicity of sensors. Additionally, the study of the sensor's data can help in optimization the design of the manipulators. In this line of thought, this paper applies two distinct methods for classification of sensors used in robotics. One of the adopted methods leads to arrange the robotic signals in terms of identical spectrum behavior. The other method is the multidimensional scaling technique applied to the correlation of the signals in the time domain. Both methods conduct to similar results, obtaining three groups of signals: the group of "positions", the group of "currents" and the group of "forces, torques and accelerations". © Springer-Verlag Berlin Heidelberg 2012.