Self organizing decentralized intelligent position controller for robot manipulator

Abstract

Our work aims at building a position controller for a two-link rigid robot manipulator via combination of decision logic methods. Decentralized self-organizing fuzzy control scheme is suggested first. The controller for each link consists of a feed forward torque-computing system and feedback PD system. The feed forward system is designed by fuzzy system and then trained and optimized offline by the intelligent methods such as adaptive neuro fuzzy inference system (ANFIS), artificial neural network (ANN) and genetic algorithm (GA), that is to say, not only the parameters but also the structure of the fuzzy system are self organized. The efficiency of these decision logic methods are compared by analyzing the performance of feed forward systems. The feed back PD system is again a fuzzy system in which proportional and derivative gains are adjusted properly to keep the closed-loop system stable. The proposed controller has the following merits: 1) It needs no exact dynamics of the robot systems and the computation is time saving because of the simple structure of the fuzzy systems. 2) The controller is insensitive to various dynamics and payload uncertainties in robot systems.