IFAC PapersOnLine 50-1 (2017) 8208–8213 ScienceDirect Available online at www.sciencedirect.com 2405-8963 © 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Peer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2017.08.1385 © 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Keywords: robust control; adaptive control; Lyapunov methods. 1. INTRODUCTION Backstepping is a control strategy that consists of stepping back towards the control input of a given system. It is a powerful tool for the design of controllers for non- linear systems in or transformable to the parameter strict- feedback form. Numerous researches have been conducted for this purpose (Li et al. (2004); Kwan and Lewis (2000)). Particularly in the case of robotic manipulator control, backstepping control has demonstrated its application (Mbede and Mvogo Ahanda (2014); Huang and Chen (2004)). But all the control strategies used in the previous listed contributions, have a drawback named ”explosion of complexity”, when the size of the system increases. Recently, two new methods have emerged to overcome this drawback: dynamic surface control (Swaroop et al. (2000)) and command filtered backstepping control (Farrell et al. (2009)). These two methods use filters to approximate the time derivative of virtual controls and therefore overcome explosion of terms problem. Dynamic surface control has been used successfully to control non-linear systems disturbed by any unwanted signals and errors (Pan and Yu (2015)). Due to its design simplicity, it was widely used to control robot manipula- tors in many situations such as perturbed case, and joints flexibility case (Huang and Chen (2004); Yoo et al. (2006)). Command filtered backstepping control is not widely used in the backstepping control of robotic manipulators be- cause of its design procedure. The design used auxiliary states in order to delete or reduce filtering errors. Few papers are found that concern robot control using com- mand filtered backstepping. In Petit et al. (2015), PD + backstepping and virtual controls are performed for vari- able stiffness robots. This paper uses command filters to cope with the problem of noisy-state measurements and to overcome explosion of terms problem. But in this valuable work, actuators electrical dynamics are not taken into account. Thus the proposed control cannot cover the wide class of robot manipulators. In fact, usually to analyse the stability of a closed loop robotic system derived by backstepping controller or prove the passivity of the closed loop robotic system, it is nec- essary to use the skew symmetry property of the matrix ˙ D(q, ˙ q) − 2C (q, ˙ q), where ˙ D(q, ˙ q) is the time derivative of inertia matrix and C (q, ˙ q) is the centrifugal and coriolis forces matrix. In the design architecture of the command filtered backstepping, auxiliary states are added to com- pensate for the filtering errors (x i+1,c − a i ), where x i+1,c is the filtered version of the variable virtual control a i . The structure of these auxiliary states render difficult the usage of the skew symmetric property. Thus, the powerfulness of the command filtered backstepping is not yet exploited to * Department of Physics, University of Yaounde I, (e-mail: josephjeanmvogo@yahoo.fr). ** Electrical and Telecommunications Engineering Department of Ecole Nationale Superieure Polytechnique, the University of Yaounde I, (e-mail: achille.melingui@yahoo.fr) *** Polytech Lille, CRIStAL, CNRS-UMR 9189, Avenue Paul Langevin, 59655 Villeneuve dAscq, France (e-mail: othman.lakhal@univ-lille1.fr) Abstract: This study derives a robust adaptive control for electrically driven robot manipu- lators using support vector regression (SVR) based command filtered adaptive backstepping approach. The robot system is supposed to be subject to model uncertainties, neglected dynamics, and external disturbances. Command filtered backstepping algorithm is extended to the case of robot manipulators. A robust term is added to the common adaptive support vector regression algorithm, to mitigate the effects of SVR approximation error on the path tracking performance. The stability analysis of the closed loop system using the Lyapunov theory permits to highlight adaptation laws and to prove that all signals of the closed loop system are bounded. Simulations show the effectiveness of the proposed control strategy. Joseph Jean-Baptiste Mvogo Ahanda * Jean Bosco Mbede ** Achille Melingui ** Bernard Essimbi Zobo * Othman Lakhal *** Rochdi Merzouki *** Robust Control for Robot Manipulators: Support Vector Regression Based Command Filtered Adaptive Backstepping Approach