The discrete-time implementation of the controllers are of importance, since almost all the implementation of controllers are done on a digital computer. Therefore, this paper attempts to provide a comprehensive treatment of neural network (NN) controller design in discrete-time for the control of a multi-input multi-output robot arm using neural networks. The NN controller exhibits learning-while-functioning-feature instead of learning-then-control and do not need the dynamics of the robotic system apriori. The structure of the NN controller is derived using filtered error notions. A uniform ultimate boundedness of the closed-loop system is given in the sense of Lyapunov. Certainty equivalence is not used, persistency of excitation is not required and regression matrix is not computed, New online tuning algorithms in discrete-time are derived, which are similar to /spl epsiv/-modification for the case of continuous-time systems, and guarantee tracking as well as bounded NN weights in nonideal situations.