摘要
针对随机系统,提出了基于多层神经网络的滤波器,并将其用于惯导初始对准中。采用BP网络替代初始对准系统中的闭环卡尔曼滤波器,可以确保系统的误差状态始终为小量,实现了惯导初始对准中的滤波与校正功能。仿真结果表明,这种方法简化了系统运算的代数结构,提高了系统状态估值运算的实时性,而对准系统的精度又与原来采用滤波器的精度相当。
Develops a filter based on a multilayer neural network for stochastic systems,which is used in the inertia navigations initial alignment.The type of BP neural network instead of the closed-loop Kalman filter in the initial alignment can keep the error small,and implement the function of estimation and alignment in the inertia navigation.This filtering structure can provide distinct advantages.It is simpler for the system's algebra structure and more attractive for real time than classical filters.Simulation results show that its precision is similar to that of the Kalman filter.
出处
《南京航空航天大学学报》
CAS
CSCD
1996年第4期487-491,共5页
Journal of Nanjing University of Aeronautics & Astronautics
基金
航空科学基金资助项目
关键词
神经网络
卡尔曼滤波
初始对准
惯性导航
neural network
Kalman filtering
original alignment
inertial navigation