机构地区: 西北工业大学
出 处: 《西北工业大学学报》 1996年第4期539-544,共6页
摘 要: 提出了一种用BP算法调整经验规则,加速倒立摆模糊控制过程的方法。并基于解析型模糊控制器的原理和人工神经网络BP算法的训练模式,对这种自适应模糊控制系统进行了仿真研究。 The introduction of BP (Back Propagation)algorithm into adaptive fuzzy systems by Wang[6] was hailed as a break-through.We believe we are the first in China,if not in the world,to accelerate fuzzy control process of inverted pendulum with BP algorithm.In our study,the introduction of BP algorithm alone is considered not enough We deem that it should be supplemented and we do supplement it with expert information already existing in the fuzzy control system. In this way, we can select appropriate initial values and insure the robustness of training.The training can be done with the help of eqs.(9).(10)and (11).In utilizing eqs.(9).(10)and(11 ),it is important to select a suitable value for training step a.If a is too large,then the error function E as given in eq.(7) may unfortunately increase and we should make a smaller. If α is too small, the training process is very slow.With the help of eqs.(3)through(6),we can obtain Information about E/u, which appears in eqs.(9) through (11),and such information is helpful in selecting a suitable value for α.Fig.5a,a simulation result, shows clearly that the slow fuzzy control process before training as indicated by curve @ is very much accelerated after training by our method as indicated by curve ②.
领 域: [自动化与计算机技术] [自动化与计算机技术] [理学] [理学]