In this paper, a SLAM-integrated MMS with the capability of efficient, consistent and robust mapping in complex environments where GNSS signal is intermittently lost is presented. The system is developed to take the feedback from mapping module to the localization module into account, which significantly improves the smoothness of the 6 DOF trajectory estimation. In particular, the complex mapping problem is modeled as a hierarchical factor graph to improve the computation efficiency of the optimization back-end. Moreover, GNSS signal is modeled as a plug-and-play constrained factor to facilitate the mapping process in complex urban environments. We evaluate the performance of the proposed method on KITTI benchmark, and compare the result with both the GNNS/INS system and SuMa, which represents the traditional MMS and the typical laser SLAM system respectively. The results confirm that the proposed method enables normal and accurate mapping even when the inavailability of GNSS is up to 60%. It outperforms the traditional MMS in terms of map consistency with an increase of 24.7% on average MME, and outperforms laser SLAM in terms of the absolute accuracy with 5.9 times increase on average RMSE.

本文提出了一种SLAM集成的MMS，它在间歇性丢失GNSS信号的复杂环境中具有高效，一致和鲁棒的映射能力。开发该系统时要考虑到从映射模块到定位模块的反馈，从而显着提高了6自由度轨迹估计的平滑度。特别地，将复杂映射问题建模为分层因子图，以提高优化后端的计算效率。此外，GNSS信号被建模为即插即用约束因素，以促进复杂城市环境中的映射过程。我们在KITTI基准上评估了该方法的性能，并将结果与​​GNNS / INS系统和SuMa进行了比较，它分别代表传统的MMS和典型的激光SLAM系统。结果证实，即使当GNSS的不可用性高达60％时，所提出的方法也能够进行正常且准确的映射。它在地图一致性方面优于传统MMS，平均MME增加24.7％，在绝对精度方面优于激光SLAM，平均RMSE增加5.9倍。