基于URTS平滑建议分布函数的PF算法设计及应用

摘要

To improve the state estimate performance of the particle filtering with the Unscented Kalman filtering proposal distribution function, the Unscented Kalman filtering and the Rauch-Tung-Striebel (RTS) fixed interval smoothing algorithm were integrated, and a new kind of proposal distribution function named as Unscented RTS smoothing proposal distribution function was designed.The Unscented RTS smoothing distribution implements the Unscented Kalman filtering, and then carries out the RTS fixed-interval smoothing to produce a prediction sample.The accuracy of prediction particles is greatly improved by the new proposal distribution function,and the state estimation accuracy of the corresponding particle filtering method (PF-URTS) is also improved.The applications of the new particle filtering algorithm to the OPS/DR integrated navigation system will prove the feasibility and validity of the proposed method.%为改善基于Unscented Kalman滤波建议分布函数的粒子滤波状态估计的性能,将Unscented Kalman滤波与RTS(Rauch-Tung-Striebel)固定区间平滑算法融合,产生了一种新的建议分布函数--Unscented RTS平滑建议分布函数.该函数首先实施Unscented Kalman滤波,之后对此滤波结果进行RTS固定区间平滑,以此产生预测样本.以此新建议分布函数得到的预测粒子的精度较通常的以Unscented Kalman滤波方法作为建议分布函数时得到的预测粒子的精度将大为提高,进而提高相应的粒子滤波算法--PF-URTS的状态估计精度.新算法的可行性和有效性在GPS/DR组合导航数据处理中得到了应用验证.