A geomagnetic SLAM method enhanced by multi-source data fusion based on smartphones

doi:10.11947/j.AGCS.2025.20250266

Abstract

Abstract:

Geomagnetic simultaneous localization and mapping (SLAM) enables smartphone-based positioning in unknown indoor environments without requiring a pre-established geomagnetic fingerprint database. However, geomagnetic SLAM on smartphones still faces technical bottlenecks, including low inertial positioning accuracy, insufficient adaptability of factor graph optimization (FGO) under dynamic conditions, and performance deterioration in large-space SLAM applications. To address these challenges, this paper proposes an enhanced optimization algorithm for geomagnetic SLAM in large-space indoor environments by designing a variance-based temporal increment mechanism and presenting multi-source data key frames. First, to enhance inertial positioning accuracy, this paper explores the characteristic patterns exhibited during the movement of pedestrians to construct observation equations, and integrate them with geomagnetic environmental information to achieve smartphone-based geomagnetic SLAM. Second, to overcome the insufficient dynamic adaptability of the FGO, a hybrid positioning framework is adopted that combines front-end Kalman filter and back-end FGO, improving timeliness. Meanwhile, a variance-based time series increasing mechanism is designed to dynamically integrate different positioning methods. Third, to alleviate the performance degradation of geomagnetic SLAM in large-space indoor environments, the concept of keyframes and their feature representation is extended along the temporal dimension, effectively alleviating the problem of large-space geomagnetic mismatching. A robust loop detection and matching algorithm is developed using multi-source data, and also a keyframe scoring mechanism is constructed to reduce spatial density and improve computational efficiency. Experimental results demonstrate that the proposed method achieves reliable geomagnetic SLAM in a large-space indoor loop-closing scene. Compared with standalone inertial positioning and classical geomagnetic SLAM, the proposed enhanced approach reduces the root mean square positioning error by 18%~67%. Moreover, it requires only 22.6% of the keyframes used by the standard approach, while still achieving higher accuracy and smoother localization results. Furthermore, experiments were conducted to investigate the impact of parameter settings on both localization accuracy and runtime, and identify the voxel grid size of basis functions as the primary factor in geomagnetic map construction.

Key words: geomagnetic SLAM, multi-source fusion, inertial positioning, factor graph optimization, extended Kalman filter

CLC Number:

P232

Kefan SHAO, Zengke LI, Meng SUN, Zhenbin LIU, Qi WU. A geomagnetic SLAM method enhanced by multi-source data fusion based on smartphones[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(10): 1812-1825.

Figures/Tables 13

Tab. 1

Fig. 1

Fig. 2

Fig. 3

Tab. 2

Fig. 4

Tab. 3

Tab. 4

Fig. 5

Fig. 6

Fig. 7

Fig. 8

Fig. 9

References 30

[1]	WANG Liang, LI Zishen, WANG Ningbo, et al. Real-time GNSS precise point positioning for low-cost smart devices[J]. GPS Solutions, 2021, 25(2): 69.
[2]	祝会忠, 孙沐凡, 李军. 灾害环境下低成本终端BDS高精度应急定位方法[J]. 武汉大学学报(信息科学版), 2025, 50(6): 1054-1064.
	ZHU Huizhong, SUN Mufan, LI Jun. High-precision emergency positioning method for low-cost terminal BDS in disaster environments[J]. Geomatics and Information Science of Wuhan University, 2025, 50(6): 1054-1064.
[3]	LIU Ao, WANG Wenguang, LI Yiming. MSPos: a geomagnetic matching suitability-incorporated positioning method in indoor environments[J]. IEEE Transactions on Instrumentation and Measurement, 2025, 74: 9507815.
[4]	陈锐志, 陈亮. 基于智能手机的室内定位技术的发展现状和挑战[J]. 测绘学报, 2017, 46(10): 1316-1326. DOI: . doi: 10.11947/j.AGCS.2017.20170383
	CHEN Ruizhi, CHEN Liang. Indoor positioning with smartphones: the state-of-the-art and the challenges[J]. Acta Geodaetica et Cartographica Sinica, 2017, 46(10): 1316-1326. DOI: . doi: 10.11947/j.AGCS.2017.20170383
[5]	LI You, HE Zhe, GAO Zhouzheng, et al. Toward robust crowdsourcing-based localization: a fingerprinting accuracy indicator enhanced wireless/magnetic/inertial integration approach[J]. IEEE Internet of Things Journal, 2019, 6(2): 3585-3600.
[6]	KUANG Jian, NIU Xiaoji, CHEN Xingeng. Robust pedestrian dead reckoning based on MEMS-IMU for smartphones[J]. Sensors, 2018, 18(5): 1391.
[7]	CHEN Changhao, LU Xiaoxuan, MARKHAM A, et al. IONet: learning to cure the curse of drift in inertial odometry[C]//Proceedings of the 32nd AAAI Conference on Artificial Intelligence. California: AAAI Press, 2018: 6468-6476.
[8]	HERATH S, YAN Hang, FURUKAWA Y. RoNIN: robust neural inertial navigation in the wild: benchmark, evaluations, & new methods[C]//Proceedings of 2020 IEEE International Conference on Robotics and Automation. New Jersey: IEEE, 2020: 3146-3152.
[9]	LIU Wenxin, CARUSOD , ILG E, et al. TLIO: tight learned inertial odometry[J]. IEEE Robotics and Automation Letters, 2020, 5(4): 5653-5660.
[10]	ANGERMANNM , ROBERTSON P. FootSLAM: pedestrian simultaneous localization and mapping without exteroceptive sensors: hitchhiking on human perception and cognition[J]. Proceedings of the IEEE, 2012, 100: 1840-1848.
[11]	SHOKRY A, ELHAMSHARY M, YOUSSEF M. DynamicSLAM: leveraging human anchors for ubiquitous low-overhead indoor localization[J]. IEEE Transactions on Mobile Computing, 2021, 20(8): 2563-2575.
[12]	李泰宇, 旷俭, 牛小骥. 基于磁强计阵列的室内行人定位算法研究[J]. 传感技术学报, 2020, 33(10): 1425-1431.
	LI Taiyu, KUANG Jian, NIU Xiaoji. Research on indoor pedestrian location algorithm based on magnetometer array[J]. Chinese Journal of Sensors and Actuators, 2020, 33(10): 1425-1431.
[13]	KOK M, SOLIN A. Scalable magnetic field SLAM in 3D using Gaussian process maps[C]//Proceedings of the 21st International Conference on Information Fusion. New Jersey: IEEE, 2018: 1353-1360.
[14]	VISET F, HELMONS R, KOK M. An extended Kalman filter for magnetic field SLAM using Gaussian process regression[J]. Sensors, 2022, 22(8): 2833.
[15]	CUI Yan, XIONG Zhi, LI Xiaodong, et al. A pedestrian SLAM scheme grounded in inertial-based map and magnetic field map[J]. IEEE Sensors Journal, 2024, 24(5): 6500-6514.
[16]	HANLEY D, DE OLIVEIRA A S D, ZHANG Xiangyuan, et al. The impact of height on indoor positioning with magnetic fields[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 8501719.
[17]	乔治. 视距-磁场SLAM融合技术研究[D]. 哈尔滨: 哈尔滨工业大学, 2021.
	QIAO Zhi. Research on visual-magnetic field SLAM fusion technology[D]. Harbin: Harbin Institute of Technology, 2021.
[18]	KOK M, SOLIN A. Online one-dimensional magnetic field SLAM with loop-closure detection[C]//Proceedings of 2024 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI). New Jersey: IEEE, 2024: 1-7.
[19]	张小红, 张元泰, 朱锋. 城市复杂场景下GNSS定位的因子图优化方法及其抗差性能分析[J]. 武汉大学学报(信息科学版), 2023, 48(7): 1050-1057.
	ZHANG Xiaohong, ZHANG Yuantai, ZHU Feng. Factor graph optimization for urban environment GNSS positioning and robust performance analysis[J]. Geomatics and Information Science of Wuhan University, 2023, 48(7): 1050-1057.
[20]	LIAO Wei, SHANG Junna, LI Zehua, et al. A PDR and magnetic fingerprint fusion positioning based on local covariance constrained multi-objective optimization[J]. Measurement, 2025, 253: 117557.
[21]	蔺小虎, 杨鑫, 姚顽强, 等. 面向复杂地下空间的多源传感器数据动态加权融合SLAM方法[J]. 测绘学报, 2025, 54(3): 523-535. DOI: . doi: 10.11947/j.AGCS.2025.20230586
	LIN Xiaohu, YANG Xin, YAO Wanqiang, et al. A dynamic weighted fusion SLAM method using multi-source sensor data in complex underground spaces[J]. Acta Geodaetica et Cartographica Sinica: 2025, 54(3): 523-535. DOI: . doi: 10.11947/j.AGCS.2025.20230586
[22]	马杰, 王雯, 邓辰龙, 等. 动态观测权调整的GNSS/INS抗差因子图优化算法[J/OL]. 武汉大学学报(信息科学版), [2025-04-30]. https://link.cnki.net/doi/10.13203/j.whugis20240493.
	MA Jie, WANG Wen, DENG Chenlong, et al. A robust factor graph optimization algorithm with dynamic observation weight adjusting[J/OL]. Geomatics and Information Science of Wuhan University, [2025-04-30]. https://link.cnki.net/doi/10.13203/j.whugis20240493.
[23]	陈锐志, 郭光毅, 叶锋, 等. 智能手机音频信号与MEMS传感器的紧耦合室内定位方法[J]. 测绘学报, 2021, 50(2): 143-152. DOI: . doi: 10.11947/j.AGCS.2021.20200551
	CHEN Ruizhi, GUO Guangyi, YE Feng, et al. Tightly-coupled integration of acoustic signal and MEMS sensors on smartphones for indoor positioning[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50(2): 143-152. DOI: . doi: 10.11947/j.AGCS.2021.20200551
[24]	WAHLSTROM N, KOK M, SCHON T B, et al. Modeling magnetic fields using Gaussian processes[C]//Proceedings of 2013 IEEE International Conference on Acoustics, Speech and Signal Processing. Vancouver: IEEE, 2013: 3522-3526.
[25]	SOLIN A, SÄRKKÄ S. Hilbert space methods for reduced-rank Gaussian process regression[J]. Statistics and Computing, 2020, 30(2): 419-446.
[26]	MAHONY R, HAMEL T, PFLIMLIN J M. Complementary filter design on the special orthogonal groupS0(3)[C]//Proceedings of the 44th IEEE Conference on Decision Control/European Control Conference (CCD-ECC). New Jersey: IEEE, 2005: 1477-1484.
[27]	SKOG I, HANDEL P, NILSSON J O, et al. Zero-velocity detection-an algorithm evaluation[J]. IEEE Transactions on Biomedical Engineering, 2010, 57(11): 2657-2666.
[28]	SHAO Kefan, LI Zengke, WANG Zhaoyuan, et al. Magnetic matching/PDR combined positioning method under the constraint of error ellipse[C]//Proceedings of the 13th China Satellite Navigation Conference. Singapore: Springer, 2022: 352-362.
[29]	GUTIÉRREZ J D, JIMÉNEZ A R, SECO F, et al. GetSensorData: an extensible Android-based application for multi-sensor data registration[J]. SoftwareX, 2022, 19: 101186.
[30]	SHU Mingcong, CHEN Guoliang, ZHANG Zhenghua, et al. Indoor geomagnetic positioning using direction-aware multiscale recurrent neural networks[J]. IEEE Sensors Journal, 2023, 23(3): 3321-3333.

使用设备	不同SLAM方案	定位范围	状态估计方法
足部IMU	惯性SLAM^[10]	28 m×20 m	RBPF
智能手机	语义SLAM与机会相遇信号^[11]	3000 m²	粒子滤波与EKF
智能手机	视觉惯性里程计辅助的地磁SLAM^[13]	40 m×50 m	RBPF
足部IMU	地磁SLAM^[14]	15 m×20 m	EKF
智能手机	惯性SLAM与地磁SLAM^[15]	20 m×40 m	RBPF
智能手机	地磁SLAM与视觉SLAM^[17]	10 m×12 m	位姿图优化
智能平板	视觉惯性里程计辅助的地磁SLAM^[18]	15 m×20 m	EKF

手机型号	指标	方法1	方法2	方法3	方法4	方法5	方法6	方法7	方法8	方法9	方法10
小米8	RMSE	6.83	8.49	7.11	7.13	6.95	6.95	8.82	13.62	5.38	3.45
	MAXE	13.62	16.76	14.17	14.20	15.74	14.16	20.49	31.64	11.14	7.20
	MEANE	5.96	7.42	6.22	6.24	6.01	6.05	7.20	11.36	4.62	3.09
红米K30	RMSE	6.88	6.73	6.87	6.88	6.29	6.08	6.19	7.41	10.47	3.74
	MAXE	13.23	13.09	13.17	13.19	12.24	12.01	11.66	16.84	21.83	7.04
	MEANE	6.08	5.85	6.08	6.09	5.57	5.31	5.37	6.33	8.89	3.35

手机型号	指标	方法1	方法2	方法6	方法10
小米8	RMSE	3.58	3.54	3.87	2.87
	MAXE	8.11	6.93	6.61	4.91
	MEANE	3.10	3.26	3.51	2.65
红米K30	RMSE	3.26	3.12	2.95	2.07
	MAXE	5.84	6.41	6.15	6.66
	MEANE	2.91	2.79	2.36	1.62

方法	RMSE	MAXE	MEANE
8	3.10	5.67	2.80
10	2.87	4.91	2.65