In this paper, a novel observer-based adaptive fuzzy-neural network (FNN) control scheme for robotic systems is proposed for tracking performance and to suppress the effects caused by uncertainties, and disturbances. A PSO-SA based adaptive FNN system is used to approximate an unknown system from the manipulation of the model following tracking errors. The proposed scheme uses an observer, which allows for identifying the state of an unknown state in the system, simultaneously. It is shown that the proposed control scheme can guarantee the better tracking performance and suppress internal uncertainties or external disturbance. Simulations are given to show the validity and confirm the performance of the proposed scheme.