| Summary: | 碩士 === 中華大學 === 工業工程與管理研究所 === 87 === ABSTRACT
Servo robots are widely applied in an industrial environment. The control of servo robot is more complicated than nonservo robot. The robot control is a key to industrial automation and in this research the servo control of a robot. The optimal control based on a performance index and real world constraints is attempted in this study. The optimal design of control system for a MISUBISHI MOVEMASTER Robot is illustrated. MISUBISHI MOVEMASTER Robot is a limited capacity educational robot.
The design of control system has been attempted for minimization of error function of angular displacement. By using Parseval's theorem, the performance index for error function has been developed for MISUBISHI MOVEMASTER Robot. Then the torque and stability constraints are formulated. The performance function under the set of constraints is then optimized through a non-linear programming technique. It is obvious that the optimal design mathematical model is inherently non-linear for robot and any linear approach will not work.
In real time implementation of optimal compensation, the technique adopted in this research offer a clear insight into the pole placement problem. It is seen in this research that if artificially large bounds are imposed on compensatory parameters, then the torque and stability constraints are active at the optimum. It has also been experienced that choice of initial x is important for the convergence and solution as it is used to form the performance index.
The MISUBISHI MOVEMASTER Robot has five joints, i,e. five degree of freedom, with a gripper. The control system designed through the non-linear optimization compares well with the available date and control parameters of the robot manufacturer. The approach has been presented with an example and the relationship of time scaling parameters versus the optimum proportional gain (kp), the optimum integral gain (ki), and the optimum derivative gain (kD) are plotted. The overall conclusion is that as the scaling parameters increases so the kp、ki and kD values are decreases before becoming flat.
|
|---|
