M. S. Thesis Abstract
Adaptive Control of a Three-Link In-Parallel Manipulator
Several adaptive control strategies were used to reduce the trajectory error of a three link in-parallel manipulator. The first controller used was a model referenced adaptive controller (MRAC) with a zero phase error tracking control (ZPETC) pre-filter. The other controllers used were a model referenced integral controller with a ZPETC pre-filter and a hybrid digital-analog controller with an adaptive ZPETC pre-filter. Also, the endpoint trajectory was measured and compared to the desired trajectory.
The MRAC adjusted the gains of a digital PD controller to make the plant respond as desired; it did not work well due to static and Coulomb friction. For the next two controllers, the plant was changed to an analog velocity servo with a high forward gain. This nearly eliminated effects due to friction.
The model reference integral controller used an integral term to reduce tracking error due to modeling error and disturbances. This controller is similar to a signal synthesis controller Yuan implemented . This controller used fixed gains on the integrator; it cannot be called adaptive in the strictest sense. However, this controller provided a good comparison to see if adaptive control was really necessary. This controller worked well; the controller did have small amounts of overshoot though.
The adaptive ZPETC used a recursive least squares algorithm to estimate system parameters in real time. Judicious choice of the plant's digital and analog feedback gains permitted modeling the plant as a first order system. This allowed using a simpler prefilter and reduced the persistency of excitation requirement for parameter convergence. This controller worked very well; the tracking error was reduced to the level of sensor noise.
Measurement of the endpoint trajectory with no payload revealed constructional error in the manipulator. Also, when a horizontal load was applied to the endpoint, manipulator compliance caused large trajectory errors.