An improved design of RBF neural network control algorithm based on spintronic memristor crossbar array

Spintronic memristors have received significant attention as a potential building block for control systems, and particularly, it can be laid out in a high-density grid known as a crossbar array. Meanwhile, radial basis function (RBF) neural network control algorithm can effectively improve the control performance against large uncertain systems. But choosing reasonable parameters for RBF neural network to save convergence time and reduce the amplitude of the initial step is very important and difficult. Therefore, based on the study of characteristics of RBF neural network and spintronic memristors, this paper proposes a RBF neural network control algorithm based on spintronic memristor crossbar array; that is to say, the RBF neural network is divided into two parts of hardware and software. Hardware completes the updated parameters’ storage and extraction by spintronic memristor crossbar array, and software accomplishes the parameters updating process from the traditional RBF neural network. With a multiple-input multiple-output data sample training as the research object, simulation results show that the typical RBF neural network control algorithm needs 24 steps to fulfill the convergence requirements and the amplitude of the initial step is close to 2e−4. However, the proposed algorithm only needs 18 steps and the amplitude of the initial step is just about 8.6e−9. The comparison of these results indicates evidently the effectiveness of proposed algorithm.