论文部分内容阅读
只用于目标方位跟踪的广义卡尔曼滤波器在理论上已得到了分析。本文介绍了状态矢量閉合式表达式和它的辅助协方差矩阵,接着用其来说明了方位和距离估值误差怎样互相影响而引起滤波器不稳定[即协方差过早地破坏(collapse)和发散]。进一步的研究指出:一般的安置初始值的技术经常要助长上述异常状态产生。这些结果具有特殊的意义,并且目前还沒有发挥其全部的优越性。具体地说,这些结果假定,使用可以供选择的安置初始值和重新线性化方法,就可以使滤波器的稳定性得到显著的改进。本文建議和讨论了一些可选择的方法。
The generalized Kalman filter, which is only used for the tracking of target position, has been theoretically analyzed. In this paper, the state vector closed-form expression and its auxiliary covariance matrix are introduced, and then used to illustrate how azimuth and distance estimation errors affect each other and cause filter instability [ie, the covariance is prematurely collapsed and Divergent]. Further research indicates that the general technique of setting initial values often encourages the generation of the above anomalies. These results have special significance and have not yet achieved their full potential. In particular, these results assume that the stability of the filter can be significantly improved using alternative initial settings and re-linearization methods. This article suggests and discusses some alternatives.