基于神经网络信息融合技术的卫星定姿算法  

Sensor Fusion with Artificial Neural Network in Satellite Attitude Determination System

作　　者：蔡琳[1] 陈家斌[1] 吕少麟[1] 丁露[1] 宋春雷[1] 

机构地区：[1]北京理工大学信息科学技术学院,北京100081

出　　处：《火力与指挥控制》2009年第2期36-39,42,共5页Fire Control & Command Control

基　　金：武器装备预研基金项目(9140A09030606BQ01)

摘　　要：姿态确定系统是卫星姿态控制系统中的重要组成部分,卫星姿态确定的精度直接影响卫星控制精度。为得到高姿态精度,针对由惯性测量单元(Inertial Measurement Unit)、红外地平仪和太阳敏感器组成的卫星姿态确定系统,分别采用BP网络算法和径向基(RBF)网络算法对不同的姿态敏感器的输出数据进行融合,并用STK(Satellite Tool Kit)数据进行了仿真。仿真分析结果表明这两种学习算法均可以提高卫星定姿精度,相对而言,RBF网络无论是精度上还是收敛速度上均优于BP网络。Attitude Determination System （ADS） is one of the most important part in satellite attitude control system, the accuracy of ADS directly affects the accuracy of the control system. The fusion of multiple attitude sensors for ADS which consists of inertial measurement unit （IMU）, horizon sensors and sun sensors, is performed via neural network architectures, back-propagation （BP） algorithm and Radial Basis Function （RBF） network are adopted separately, and the simulation is processed using the STK （Satellite Tool Kit） data in the present study. Simulation results indicate that the accuracy of the system is improved by the two proposed methods, by comparison, both the accuracy and convergence speed of RBF network are better than BP network.

关 键 词：神经网络 信息融合 卫星姿态确定 姿态敏感器 

分 类 号：V448.22[航空宇航科学与技术—飞行器设计]