Learning control of robot manipulators

摘要

Recent interests in intelligent approaches to robotics and manufacturing have resulted in great promise for efficient use of robots in production processes. Although neural-driven robotic controllers are still some years away, expert systems have, on the other hand, proven to be an implementable approach for many off-line applications. In this paper, a rule-based expert learning controller is introduced to control a multi-link robot manipulator. The proposed control approach has a proven convergence through two-dimensional (2-D) systems theory and is very robust with respect to the variations of the robot's dynamic model. Simulation results for a 2-link robot show very promising results.