基于无序量测粒子滤波的无人船导航  

Unmanned ship navigation based on out-of-sequencemeasurement particle filter

作　　者：古毅杰 张闯[1] 康凯航 GU Yijie;ZHANG Chuang;KANG Kaihang(Navigation College,Dalian Maritime University,Dalian 116026,Liaoning,China)

机构地区：[1]大连海事大学航海学院,辽宁大连116026

出　　处：《上海海事大学学报》2022年第4期9-15,共7页Journal of Shanghai Maritime University

基　　金：国家自然科学基金(51879027)。

摘　　要：针对粒子滤波(particle filter,PF)的局部量测模糊问题,提出一种基于无序量测的模糊更新方法,并将其应用于无人船的组合导航中。根据模糊量测更新的观测量与滤波器估计的先验分布之间的耦合关系,求取基于序贯重要性重采样的PF框架的无序量测解;当量测量模糊或不足时,跳过此时间步长对应的量测更新,用无序量测解弥补跳过的量测更新,进而以较低的协方差获得准确的位置估计。通过无人船导航实验验证该方法的有效性。与其他常用的滤波方法相比,该方法在位置估计均方根误差指标上的表现更优。Aiming at the fuzzy problem of local measurement in the particle filter(PF),a fuzzy update method based on the out-of-sequence measurement is proposed and applied to the integrated navigation for unmanned ships.According to the coupling relationship between the observation values of fuzzy measurement update and the priori distribution of filter estimation,the out-of-sequence measurement solution of the PF framework based on the sequential importance resampling is obtained;when a measurement value is fuzzy or insufficient,the measurement update corresponding to this time step is skipped,and the skipped measurement update is compensated by the out-of-sequence measurement solution,so as to achieve the accurate position estimation with low covariance.The effectiveness of this method is verified through the unmanned ship navigation experiment.Compared with other commonly used filtering methods,this method has better performance in the root mean square error index of position estimation.

关 键 词：模糊更新 无序量测 粒子滤波(PF) 组合导航 

分 类 号：U666.11[交通运输工程—船舶及航道工程]