进化规划改进算法及其在舰载导航中的应用  

Improved evolutionary programming and its application in navigation system

作　　者：刘建娟[1] 徐晓苏[2] 

机构地区：[1]河南工业大学电气工程学院,郑州450007 [2]东南大学仪器科学与工程学院,南京210096

出　　处：《大连海事大学学报》2010年第2期67-70,74,共5页Journal of Dalian Maritime University

基　　金：国防973项目(973-61334);国家自然科学基金资助项目(50575042);河南省教育厅自然科学研究计划项目(2008BS120003);河南工业大学博士基金资助项目(2007BS049)

摘　　要：为解决设计BP神经网络时所遇到的网络参数和连接权重难以确定,在随机扰动下不能达到最佳效果,学习时间较长难以满足系统实时性要求等问题,引入进化神经网络.根据舰载导航的要求及特点,对基于进化规划的BP神经网络进行设计,并将之应用于舰载导航系统中.仿真结果证明,该方法加快了神经网络的收敛速度,解决了BP神经网络存在的问题.并在舰载导航系统外观测数据不可得时,规避卡尔曼滤波所存在的问题,从而保证了卡尔曼滤波器的正常运行,进一步提高了舰载导航系统的精度.To solve such problems as undeterminable network parameters and weights,unacceptable effect under random perturbation,and long learning time which not meeting the real-time requirement of system when designing BP neural network, the evolutionary neural network was introduced.According to the requirements and characteristics of the navigation system,BP neural network was redesigned based on evolutionary programming,and then it was used in integrated navigation system.Experimental results indicate that the proposed method can accelerate convergence speed of the neural network,which make up the deficiencies of the BP neural network.The proposed method avoids the problems of Kalman filter when the outside observation data is unreliable,which ensures the normal operation of Kalman filter with a higher accuracy,thereby the accuracy of the navigation system is improved.

关 键 词：惯性导航 卡尔曼滤波 小生境技术 进化规划 进化神经网络 

分 类 号：TP273.2[自动化与计算机技术—检测技术与自动化装置]