在对现有的抗差滤波方法作进一步分析的基础上,针对观测模型和动力学模型同时存在误差的情况,提出了一种新的抗差滤波——分步抗差自适应滤波。该方法在滤波之前,利用抗差岭估计等抗差有偏估计得到参数估计值,并修正观测向量,把观测向量中的粗差控制在可接受范围之内,然后在滤波中利用自适应因子消除动力学模型误差的影响,这样分两步分别对观测模型误差和动力学模型误差进行控制。大量的GPS动态导航定位试验验证了本文方法的有效性和稳定性。
A robust and adaptive filtering algorithm is proposed. Firstly, the observed values are corrected by the robust biased estimation such as robust ridge estimator, which only uses the original observed information before the calculation of the filter. Secondly, the errors of the dynamic model are eliminated through the adaptive factor. The effects of the errors from the measurement model and dynamic model can be resisted in two steps. Finally, a large number of GPS kinematic navigation calculative experiments were given to prove the validities and stabilities of this method.