A Soft-Computing Algorithm-based Cognitive Observer for an Elastic Robotic Arm

This paper proposes the use of Support Vector Machine (SVM) algorithm for modeling and states estimation of an elastic robotic arm. Due to the complexity of the elastic robotic arm, an accurate mathematical model and large number of sensors are needed to achieve accurate estimation. Initially, it is assumed that all system states are measurable for a short period of time. Additionally, the system output and input signals are being continuously measured. The modeling module builds two models, the internal model which will capture the dynamics of the elastic robotic arm and the data-driven observer model which will estimate the system states. The internal model and data-driven observer are obtained based on the experimental measurements using SVM algorithm in contrast to classical methods that uses the mathematical system model to drive an observer. The internal model is able to make multi-steps ahead predictions of all system states; therefore it can be used to generate a suitable control strategy. Once the models are ready, the main sensors can be removed or turned off. The proposed modeling module will eliminate the need for mathematical modeling and reduces the number of permanent sensors needed. If the main sensors are removed completely, the hardware price can be radically reduced. The simulation result demonstrates the efficiency and high performance of the modeling module. (© 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)