Comparative evaluation of real time robot control systems

摘要

This thesis studies the problem of the optimal performance of real-time controllers for robotic manipulators. The robotic system under study is an Hirata AR-350i SCARA robot which is a small assembly robot. This robot has the advantage that it can be used safely with a single operator and has a well known dynamic model. Experimental results are obtained using the physical robot and using a thoroughly verified computer simulation of the robot based on experimentally derived dynamic model parameters. There are a number of important issues to resolve in such a study. One issue is the question of how to quantify controller performance using the most effective and discriminating technique. In this thesis a survey is conducted that compares and contrasts classical control theory metrics with composite measures based on the system response parameters. The outcome of this survey shows that the same trend information is obtained with both measures therefore the pragmatic decision was to use the measure that was most readily calculable. This is the integral-error-squared term. A further important issue for the study of robotic manipulator performance is the choice of controller. In this research a number of previously published and commonly used controllers are selected for implementation on both the physical robot and in the simulation environment. These controllers include the widely used PID controller as well as model based controllers such as the Computed Torque Method (CTM) controller, the Model Reference Adaptive (MRAC) controller and the Variable Structure (VSC) controller. The research is extended beyond the use of existing controllers by the development of a novel hybrid controller. This controller exploits the advantages of the CTM controller and the VSC controller by combining components of these two controllers.