上面已指出,Wiener和kalman滤波技术都要求掌握协方差矩阵Q和R的完整信息。Kalman法还要求知道误差方差矩阵的初始值。事实上,这些先验消息一般是不适用的。例如,在确定深空间中飞行器轨道的问题方面,一般是以Doppler形式进行观测的。计算的Doppler和距离数据均含有由随机起伏、离子层扰动、接收机噪声和计数器量化噪声等构成的观测噪声。此外,飞行器还受到太阳的作用、流星群的碰撞和在运行到行星途中(到火星约需200天)燃料泄漏等各种扰动。要精确地确定任一种干扰的统计特性都有困难。由于飞行器喷射条件的不确定性,在中程机动之后,座际数据进入滤波器,且在估值的初始阶段内,严重地影响滤波 It has been pointed out above that both Wiener and kalman filtering techniques require the completeness of the covariance matrix Q and R. The Kalman method also requires knowing the initial value of the error variance matrix. In fact, these a priori messages are generally not applicable. For example, Doppler observations are generally used in determining the orbit of aircraft in deep space. The calculated Doppler and distance data all contain observed noises consisting of random fluctuations, ionospheric disturbances, receiver noise and counter quantization noise. In addition, the aircraft is also affected by the sun, the collision of meteors and various disturbances such as fuel leaks on its way to the planet (about 200 days to Mars). It is difficult to accurately determine the statistical characteristics of any kind of interference. Due to the uncertainty of aircraft jetting conditions, inter-vehicle data enters the filter after mid-range maneuvers and, during the initial period of the valuation, severely affects the filtering