基于忆阻神经网络PID控制器设计

Design of Memristive Neural Network PID Controller

忆阻器是一种阻值变化依赖于通过它的电荷量或磁通量的新型电子器件,在非易失性存储器和人工神经网络等应用中具有独特优势.智能PID控制器在现代控制领域中占据重要地位.文中在研究传统单神经元和神经网络PID控制器的基础上,引入具有类似于人脑突触行为的忆阻器实现新型的权值更新机制,提出了忆阻单神经元和神经网络PID控制器模型.在理论推导的基础上,进行了编程仿真和Simulink建模.实验结果验证了提出方案的有效性.该方案可望进一步推动智能PID控制器的发展,促进神经网络PID控制器的硬件实现.

Memristor is a novel kind of electronic device whose resistance can continuously change depending on the charge or flux passing through it, and has unique advantages in the applications of nonvolatile memory and artificial neural networks. The intelligent PID controllers play an essential role in modern control field. Based on the study of conventional single neuron and neural network PID controller, this paper presents an approach to the memristive weight updating which is similar to the biologic synapse, and puts forward a model of the memristive single neuron and neural network controllers. According to the stipulating theory, the program simula- tion and Simulink modeling have been already provided. Experimental results verify the effective- ness of proposed scheme. It would be beneficial for the development of the modern intelligent PID controller and expect to promote the hardware realization of the neural network PID controller.