驾驶员自适应神经模糊推理模型的建立和仿真.pdfVIP

14 5 Vol. 14 l .5 2007 10 ELECTRONICS O TICS CONTROL Oct.2007 : 1671- 637 Ú(2007) 05-0102-04 1 1 2 董文瀚, 孙秀霞, 林 岩 ( 1. , 710038; 2. , 100083) : 提出了一类基于自适应神经模糊推理系统(ANFIS) 的飞机驾驶员建模方法首先建 立了包 若干待定模糊规则参数的基本驾驶员Takag-i Sugeno 模糊模型, 将其转换成等效的ANFIS 网络系统后, 采用驾驶员实际操纵数据和自适应B 算法对模糊规则各 个待定参数进行训练并建 立最终的模型进而对已建立的驾驶员模型进行人机闭环系统仿真, 结果表明, 较之传统的传递函 数模型, ANFIS 模型与驾驶员实际操作输出之间的拟合程度更加理想 : 驾驶员模型; 模糊推理; 神经网络; 自适应控制; 飞行控制 : V271.4; T 273.4 : A Design and simulation of pilot model based on adaptive neural fuzzy inference system 1 1 2 DONG Wen-han , SUN Xiu-xia , LIN Yan ( 1. Engineering Institute, Air Force Engineering University, Xi. an 710038, China; 2. School of Automation, Beji ing University of Aeronautics an Astronautics, Beij ing 100083, China) Abstract: A method for building pilot models is proposed based on Adaptive Neural Fuzzy Inference System (ANFIS) . A basic Takagi-Sugeno fuzzy model of pilot is built at first, which contains several fuzzy-rule pa- rameters to be determined. It is then transformed into the equivalent ANFIS network system. We use the actual flight data and the adaptive Back- ropagation( B ) algorithm to train the undetermined parameters of ANFIS, and obtained the final model. The simulation results of pilot-aircraft close- loop system showed that, compar- ing with the traditional transfer function methods, the ANFIS pilot model presented by our scheme is more rea- sonable and can illustrate the handling behavior more accurately. Key words: pilot