论文部分内容阅读
本文提出了一种可直接用作系统控制器的神经元模型。该模型是一种单层感知机,是 Perceptron 算法的一种推广。大量的仿真实验表明,该自学习控制神经元不仅适用于几乎全部 SISO 线性定常系统的控制,而且还适用于部分非线性系统的控制;当系统模型改变时,无需改变算法就能保证系统高精度的稳定输出,并且具有响应速度快、鲁棒性强等优点。若与经典和现代控制理论相结合,则具有更宽的适用范围。该算法对无法获取精确数学模型的系统控制具有一定的实用价值。
This paper presents a neuron model that can be used directly as a system controller. The model is a single-layer perceptron and is a generalization of Perceptron’s algorithm. A large number of simulation experiments show that this self-learning control neuron is not only suitable for the control of almost all SISO linear steady-state systems, but also for the control of some nonlinear systems. When the system model is changed, the system precision can be guaranteed without altering the algorithm Stable output, and has the advantages of fast response, robustness and so on. If combined with classical and modern control theory, it has a wider scope of application. The algorithm has certain practical value for the system control which can not acquire the accurate mathematical model.