城市峡谷环境多源导航信息弹性融合定位方法  

Multi-source Navigation Information Resilient Fusion Method under Urban Canyon Scenario

作　　者：孟骞 姜颖颖 王立辉[1,2] 胡祎雯[1] 李圣英 MENG Qian;JIANG Ying-ying;WANG Li-hui;HU Yi-wen;LI Sheng-ying(School of Instrument Science and Engineering,Southeast University,Nanjing 210096;Key Laboratory of Micro-inertial Instruments and Advanced Navigation Technology,Ministry of Education,Southeast University,Nanjing 210096)

机构地区：[1]东南大学仪器科学与工程学院,南京210096 [2]东南大学微惯性仪表与先进导航技术教育部重点实验室,南京210096

出　　处：《导航与控制》2023年第2期16-22,共7页Navigation and Control

基　　金：国家自然科学基金(编号:62203111);测绘遥感信息工程国家重点实验室开放基金(编号:21P01);微惯性仪表与先进导航技术教育部重点实验室开放基金(编号:SEU-MIAN-202101)。

摘　　要：为了提高城市峡谷环境下组合导航系统的定位精度和可靠性,提出了一种基于观测传感器独立更新的交互式弹性融合方法。该方法将观测传感器的先验置信度以马尔科夫链的形式引入多源信息融合过程中,与贝叶斯最大似然后验估计相结合,实现对各独立传感器的权重更新以及交互式融合,最终输出多源融合导航结果。典型城市峡谷环境下的车载实验表明:动态权重具有天然的故障检测与隔离功能,能够有效应对传感器噪声与粗差的干扰,实现多传感器的在线优选,传感器独立更新结构可以随时接纳可用导航传感器实现导航源的即插即用。与传统的基于扩展卡尔曼滤波的多传感器信息融合方法相比,所提出的弹性融合方法展现出更高的定位精度和鲁棒性。To improve the positioning accuracy and reliability of multi-sensor navigation system under urban canyon scenarios,a resilient interactive fusion method based on sensor-independent-update is proposed.The trust priority of navigation sensors is introduced into the information fusion in the form of transition probability matrix defined by Markov chain.The overall estimation is determined by the weighted sum of average from every filter estimate.This weight of every model is dynamic updated by the prior transition information and posterior model likelihood.The multi-sensor integration is implemented in state estimation domain enhanced by interactive information fusion rather than in measurement domain implemented in traditional filter method.The kinematic vehicle experiment in urban canyon environment verifies the superiority of the proposed method.The dynamic weights have the natural function of fault detection and exclusion without additional operations which is insensitive to sensor measurement noise and outliers.The effect of dynamic sensor selection is achieved in this processing.The same independent structure is also applied to adopt new available sensor to realize plug-and-play navigation.The proposed resilient navigation method shows better accuracy and reliability compared to classical Kalman filters.

关 键 词：弹性导航 多源信息融合 城市峡谷 马尔科夫链 

分 类 号：U666.11[交通运输工程—船舶及航道工程]