A fuzzy normalized sliding mode learning scheme for the trajectory tracking controls of a two-linkage robot arm

A novel learning scheme for the trajectory tracking controls of a two-linkage robot arm, based on fuzzy normalized sliding mode, is presented in this paper. Learning control is a betterment process toward the desired trajectory commands with an iterative mechanism to compensate for the regular perturbations through the consecutive adjustment of control efforts and is very useful for controlling an uncertain system to have excellent transient performance. However, typical learning schemes are to focus on the stability of learning only, not efficient enough to overcome the variation of the perturbations and to handle the dynamics of learning behavior as well, so as to have the possibility to make the learning action fallen into local minimum. It has been known that sliding mode schemes with hyperplanes can constrain a controlled system toward a desired action to perform robust characteristics. Therefore, if combining the concept of sliding mode schemes with learning, the advantages of setting the convergent dynamics in advance and having robust learning behavior could be obtained directly. In addition, to further resolve the inherent drawbacks of chattering and the trouble of parameter design of sliding mode scheme, a novel approach in betterment process through a normalized sliding mode scheme to improve the learning rate and tracking errors of a robot manipulator is proposed in this paper. For validations, the trajectory controls for a two-linkage robot arm are carried out and as might have been expected, the excellent smooth learning dynamics can be achieved obviously.