重庆理工大学学报2024，Vol.38Issue(13)：1-11,11.DOI:10.3969/j.issn.1674-8425(z).2024.07.001
基于因子图优化的激光惯性SLAM方法研究
Research on method of laser inertial SLAM based on factor graph optimization
摘要
Abstract
SLAM with the integration of LiDAR and Inertial Measurement Unit is an important means of localization and mapping for autonomous driving in denial environments.It comprises front-and back-end two data processing modules.In current back-end data processing,problems including high accumulative error after long-time running,high calculated load and unsatisfactory robustness in complicated denial environments still exist.To address these issues,this paper proposes a new back-end data processing algorithm adapting to laser inertial odometer based front-end algorithm.Factor graph optimization is applied to build the framework.First,an IMUpre-integration model between laser consecutive key frames is built and employed as the optimization factor for describing the IMUdata in factor graph framework,helping reduce the calculated load of data processing.Second,an efficient loop-closing detection method based on the Scan-Context descriptor is introduced,where three-dimensional spatial structure features of the point cloud are converted into a two-dimensional feature map.The calculation efficiency,therefore,is improved in the premise of keeping an eligible loop-closing detection precision.Finally,with the integration of the front-end odometer information,the whole SLAM model is built using odometer factor,IMUpre-integration factor and loop-closing detection factor,and the optimized position-orientation is solved by nonlinear optimization algorithm.In open-source-dataset and real-vehicle-based tests,the proposed DSC-Algo method exhibits markedly higher global localization precision,better calculating performance and robustness in complicated denial environments compared with other mainstream methods including FAST-LIO2,LIO-SAM and SC-LeGO-LOAM.关键词
自动驾驶/激光惯性SLAM/后端数据处理/因子图优化Key words
autonomous driving/laser inertial SLAM/back-end data processing/factor graph optimiza-tion分类
交通工程引用本文复制引用
兰凤崇,魏一通,陈吉清,刘照麟,熊模英..基于因子图优化的激光惯性SLAM方法研究[J].重庆理工大学学报,2024,38(13):1-11,11.基金项目
国家自然科学基金项目(52175267) （52175267）
中国博士后科学基金项目(2023M740817) （2023M740817）
广东省科技计划项目(2020B1212060010) （2020B1212060010）
国家车辆事故深度调查体系项目(ZL-ZHGL-2023003) （ZL-ZHGL-2023003）
