基于模糊神经网络的5连杆双足机器人混杂控制

针对双足机器人单脚支撑期控制问题, 提出了一种新型的模糊神经网络混杂控制方法. 该种方法结合了模糊神经网络、H_∞ 控制及逆系统方法的优点. 应用了一种新的多层模糊CMAC神经网络对系统进行逼近, 一方面将模糊神经网络的构造误差看作系统的干扰, 利用H_∞ 控制对干扰进行抑制. 另一方面利用模糊神经网络对系统模型进行逼近, 为逆系统的构建和H_∞ 控制率的设计提供了有效的系统信息. 并证明了在采用本文提出的模糊神经网络和自适应算法后可以抑制 L₂ 增益.

Five-Link Biped Robot Hybrid Control via Fuzzy Neural Networks

The paper presents a new fuzzy neural networks (FNN) hybrid controller to solve the trajectory tracking problem of biped robots in the single_support phase. The advantages of fuzzy neural network, H-infinity controller and inverse system method are integrated in this paper for control purpose. A new multi-layers fuzzy CMAC is applied to approximate the system information of biped robot .On the one hand, we regard construction errors of FNN as external disturbances, and then use H-infinity controller to attenuate such disturbances. On the other hand, apply the strong approximate capability of FNN to construct the inverse system and offer efficient system information to H-infinity controller. As the result, L-2 gain can be attenuated by the presented fuzzy neural network structure and adaptive algorithm.