Adaptive Control of an Industrial Robot with Application To Arc Welding

Abstract

<p>This thesis addresses the problem of position control of an industrial robot. The robot used for this study is a modified UNIMATE-2000 with five degrees of freedom, which is dedicated to research in arc welding applications.</p> <p>The problem of position control of an industrial robot is first analysed. It is evident that due to several factors such as the elasticity of the arm, gravitational effects due to link orientation and load variations, a conventional fixed feedback controller is not adequate. It is therefore necessary to use an adaptive control scheme for position control.</p> <p>It is well known that the dynamics of a robot are non-linear and coupled, because of which it is not convenient to derive control laws which can be implemented in real time. Further, because the dynamics change as a function of link orientation and load variation, the model has to be evaluated on-line.</p> <p>In the past, robots were modelled from lays of classical mechanics. Although this scheme can result in accurate models, they are not suitable for adaptive control applications because of their complexity. In addition, this scheme of modeling requires a-priori knowledge of mass of links, specification of servo amplifiers and actuators. Often, to simplify models, dynamics of actuators and elasticity of links is neglected which results in unsatisfactory control.</p> <p>In this thesis we describe a new method for modeling the robot. The model is derived from experimental observations and does not require knowledge of the structure and internal subsystems of the robot. Further, it includes actuator dynamics and the elasticity of the arm. The method is demonstrated by modeling one axis of the UNIMATE-2000.</p> <p>Since the model is of non-minimum phase, the explicit pole-placement type of a self-tuning regulator was designed. Parameters of the model are updated at every sampling interval using a recursive Ieast squares estimator. Since, as mentioned before, dynamics of the robot change as a function of link orientation and Ioad variation, a 'weighting factor' is used for parameter estimation. Studies presented include the effect of truncation of controller parameters, adaptivity of the regulator to changes in system dynamics, and the effect of a variable weighting factor.</p> <p>Finally, several aspects of implementation of the self-tuning regulator, including selection of the controller structure, selection of the microprocessor, AID, D/A converter and programming language are discussed.</p>