基于滑动窗因子图优化的多源导航信息融合  

Multi-Source Navigation Information Fusion Based on Sliding Window Factor Graph Optimization

作　　者：宋丽君[1] 赵万良 成宇翔 张雷 崔超 王鑫 SONG Lijun;ZHAO Wanliang;CHENG Yuxiang;ZHANG Lei;CUI Chao;WANG Xin(School of Information and Control Engineering,Xi'an University of Architecture and Technology,Xi'an 710055;Shanghai Aerospace Control Technology Institute,Shanghai 201109;School of Transportation Engineering,Tongji University,Shanghai 200092;School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200030)

机构地区：[1]西安建筑科技大学信息与控制工程学院,西安710055 [2]上海航天控制技术研究所,上海201109 [3]同济大学交通运输工程学院,上海200092 [4]上海交通大学电子信息与电气工程学院,上海200030

出　　处：《飞控与探测》2024年第5期20-29,共10页Flight Control & Detection

摘　　要：自动驾驶作为未来汽车产业的演进方向,其核心技术自主导航具有至关重要的地位。基于多传感器数据融合,采用滑动窗因子图优化对自主导航技术展开研究。前端通过融合惯性测量单元(Inertia Measurement Unit,IMU)与激光雷达(Light Detection and Ranging,LIDAR),输出车辆位姿以及局部点云地图,算法后端在IMU预积分和LIDAR里程计约束的基础上添加全球导航卫星系统(Global Navigation Satellite System,GNSS)因子与回环检测因子,使用滑动窗因子图优化算法进行多源导航信息融合,有效地对前端的累积误差进行修正,提升了算法精度及鲁棒性。跑车实验结果表明,当导航系统遭遇传感器故障时,运用滑动窗因子图理论构建的多源导航信息融合算法稳定,基本不受传感器故障影响,显著增强了导航系统可扩展能力。As the direction of the automobile industry,autonomous driving holds a crucial position,particularly in autonomous navigation.The paper investigates autonomous navigation technology which use a sliding window of factor graph optimization based on IMU/LIDAR/GNSS multi-source navigation information fusion.The front end integrates an inertia measurement unit(IMU)and Light Detection and Ranging(LIDAR)to output vehicle attitude and local point nephogram.Based on IMU pre-integration and LIDAR odometer constraints,the algorithm back-end adds Global Navigation Satellite System(GNSS)factors and loop detection factors.The sliding window of the factor graph optimization algorithm is used for multi-source navigation information fusion,which effectively corrects the cumulative error of the front-end and improve the accuracy and robustness of the algorithm.The results of the car experiment show that the algorithm of multi-source navigation information fusion which is based on the sliding window of the factor graph is stable and unaffected by sensor failures when the navigation system encounters sensor failures,and the method is significantly enhanced the scalability of the navigation system.

关 键 词：因子图 滑动窗 多源导航信息 惯性测量单元 SLAM 

分 类 号：U666.1[交通运输工程—船舶及航道工程]