A&QT-R 2004 (Theta14) 2004 IEEE-TTTC - International Conference on Automation, Quality and Testing, Robotics May 13-15, 2004, Cluj-Napoca, Romania Model Based Robot Control with Friction and Payload Estimation L˝ orinc M´ arton Dept. of Electrical Engineering Sapientia Hungarian University of Transylvania 540053 Tg. Mures P-ta. Trandafirilor 61. Romania martonl@ms.sapientia.ro Abstract The present study deals with control of robotic systems with unknown friction parameters and payload mass. A tracking control algorithm which uses the model of the robotic arm combined with adaptive parameter estima- tion techniques were developed to solve the proposed control problem. Using Lyapunov method it was shown that the resulting controller achieves guaran- teed final tracking accuracy. Simulation results are presented to illustrate the effectiveness and achievable control performance of the proposed scheme. Keywords - robot control, friction, parameter estimation, Lyapunov stability 1 INTRODUCTION It is well-known that nowadays the robotic applications require increased transient performances and good path tracking proprieties. To achieve these requirements the introduction of the mathematical model in the control algorithm is necessary. There are analytical methods to determine the exact mathematical model of a robot, for example using the Euler-Lagrange method [1]. The parameters of the model (masses, inertias, length of the arms) are often catalog data which are given by the manufacturer of the robotic system [2]. The friction phenomena which should be considered in every mechanical system can be described with models whose parameters are time varying, depending on external factors. These parameters cannot be determined a-priori. At the other hand there is a tendency in robotic industry to build robots with lightweight arms to avoid unnecessary energy consumption. For this reason in the robot model the mass of the payload cannot be neglected related to the masses of the arms. The mass of the payload in many application is unknown, varies according to the specific task of the robot. These consideration suggests that the friction parameters and the mass of the payload should be estimated on-line. If only the friction forces and the payload are unknown in the model it is unnecessary to use adaptive control algorithms which estimate all the parameters of the robot. In the past years the adaptive control methods [3] became a wide spread method to control not only the linear but also the nonlinear systems with unknown param- eters. Adaptive control of robotic manipulators has also been intensively studied in the control community. Early results can be found in the work of Slotine and Li [4]. The introduction of a neural network in the adaptive control law to handle 1 of 6