Self-tuning control of robot manipulators

Abstract

Adaptive control techniques are becoming increasingly important in the field of robot control in order to maintain consistency and accuracy in the face of changing robot dynamics particularly in applications involving unstructured environments, high speeds or lightweight manipulators. Several adaptive schemes are available but it is very difficult to provide theoretical analysis for their performance within a robot environment. Furthermore many adaptive schemes involve large amounts of computation which makes them difficult to implement economically using current computer technology. However this paper presents a scheme which satisfies cost constraints whilst providing high control rates. This is achieved using a combination of simple control structure, a high efficiency adaption algorithm, a low cost microprocessor together with assembly level coding and limited precision floating point arithmetic. The resultant controller provides computation times down to 2ms per adaption/control cycle, and has shown significant improvements over the fixed parameter equivalent, when compared on a simulation of an industrial robot.