摘要
经典的卡尔曼滤波器要求假设系统的动态模型和观测模型的噪声统计特性已知,而组合导航系统的噪声具有非先验性。为了解决这一问题,提出了一种新型复合神经网络(CNN)辅助卡尔曼滤波器(Kalman)。仿真试验结果表明该辅助算法的精度与一般卡尔曼算法相比提高了2倍,收敛时间缩短近200s,并有效地克服了传统神经网络学习速度慢、泛化能力弱的缺点,使系统具有自适应能力以应付动态环境的扰动。
The conventional Kalman filter assumes that the statistical properties of the noise in dynamic model and observation system are exactly known, but the noise in integrated system is uncertain. So the paper puts forward a new method using combined neural network-aided Kalman filter. Simulations suggest that the precision of complex neural network (CNN) is 2 times more than that of conventional Kalman filter, and the convergence time is 200 s less than the latter. Thus it can overcome the shortcomings of conventional neural network, such as slow learning speed and poor generalization ability, and the whole system has adaptive capability to deal with the disturbance in dynamic situation.
出处
《中国惯性技术学报》
EI
CSCD
2005年第1期50-53,共4页
Journal of Chinese Inertial Technology
基金
国家杰出青年科学基金项目资助(40125013)
