A major objective in this paper is the application of highly efficient connectionist architectures for fast and robust learning of dynamic relations used in robot control at the executive hierarchical level. Two types of neural network control structures are considered: a single-layer neural network and a multilayer perceptron. The proposed connectionist learning models are applied as a form of intelligent feed-forward robot control in the frame of decentralized control algorithm with feedback—error learning method. The Final result of this approach is a trainable robot controller with excellent learning properties. Efficiency and verification; of the proposed algorithms through simulation examples of robot trajectory tracking is shown.