机械臂的自旋忆阻RBF神经网络控制系统设计

Design of RBF neural network control system based on spintronic memristor for robotic manipulator

利用自旋忆阻器和径向基函数（RBF）神经网络融合技术为参数完全未知的机械臂在位置和速度跟踪方面设计了一种新的自旋忆阻RBF神经网络控制算法。由于自旋忆阻器可作为电子突触为神经网络提供神经元之间的连接并存储信息,且其尺寸与大脑突触在同一数量级,利用＂软件硬化＂的设计思想,将忆阻器网络（交叉阵列）忆阻值的变化替代RBF神经网络的权值更新,可以为后续的控制过程减少冗余而繁琐的训练过程,节省了时间和能量。并以二关节机械臂为控制对象进行仿真实验,实验结果验证了所提控制算法的正确性和有效性。

A new spintronic memristor neural network control algorithm is designed using the fusion technology of spintronic memristor and radial basis function（ RBF） neural network for the position and velocity tracking of the robotic manipulator with completely unknown parameters. Because the spintronic memristor can be used as electronic synaptic to connect the neurons in the neural network and store information and it has the size in the same order of magnitude with the brain synapses,using the design idea of ＂software implemented by hardware＂,the memristance variability of spintronic memristor network（ crossbar array） can be used to replace the weight update of RBF neural network,which can reduce the redundant and tedious training process and save time and energy for the subsequent control process. A two-link robotic manipulator was taken as the control object to conduct simulation experiment,the simulation results verify the correctness and effectiveness of the proposed RBF neural network control algorithm based on spintronic memristor.