A robust adaptive neural tracking controller is proposed for a class of MIMO non-linear systems with strongly coupled interconnections. With the help of RBF neural networks (NN) as approximator, a unified and systematic procedure is developed by fusion of 'dynamic surface control (DSC)' with 'minimal learning parameters' algorithm. As a result, both problems of 'explosion of complexity' and 'curse of dimension' are solved synchronously, especially, the number of updated parameters for each subsystem is reduced to two, which can reduce the computational burden to an extreme extent, consequently ease the implementation of the algorithm. Additionally, the possible controller singularity problem can be removed and the stability of the closed-loop system is guaranteed. Simulation results validate the effectiveness of the proposed scheme.