| Summary: | 碩士 === 國立成功大學 === 工程科學系碩博士班 === 100 === Both optimal control and predictive control are used and compared to control the model of a robot arm. The main reason for using the predictive control approach is that we want to attain the optimal performance by predicting the force/torque required to track a desired output trajectory. Another reason is that the predictive control approach enables controller design to be fulfilled in real time by system identification without knowing the system model.
The moving motions of a robot arm can be roughly classified into the swinging motion and the raising motion, and their movement patterns resemble the movement of a pendulum and an inverted pendulum, respectively. Thus we resort to the models of a pendulum and an inverted pendulum for deriving the dynamic equations and to compare the control results of the optimal control approach and the predictive control approach. We then apply the two control approaches to a three-link robot arm. For predictive control, we utilize the input-output data from a nonlinear system to perform system identification and acquire the linear time-invariant system to control our model. For optimal control, we use the linear quadratic regulator design on a linearized model for small motion. We derive the arm trajectories from simulated results for angles, as done in geometric kinematics. All simulations of robot arms are performed on time intervals of 1 millisecond. A comparison of these two trajectories shows that the predictive control approach is superior to the optimal control approach in reaching our desired position in finite time.
|
|---|
