SA-LM Optimization-based Fusion Positioning Method Using GNSS/Cellular Signals of Opportunity
-
摘要: 在全球导航卫星系统(GNSS)可见卫星数量不足或观测几何构型退化的受限环境下, 针对GNSS/蜂窝机会信号融合定位中观测质量差异大、非线性优化易陷入不利局部极值甚至发散的问题, 提出一种基于模拟退火Levenberg-Marquardt优化的融合定位方法. 首先, 在迭代扩展卡尔曼滤波状态估计框架下, 将模拟退火机制引入Levenberg-Marquardt优化过程, 对未满足确定性接受条件的候选步采用概率接受准则, 以提高退化观测条件下的迭代收敛可靠性, 并给出算法的收敛性分析. 其次, 构建由定位精度因子与伪距残差驱动的双因子动态加权机制, 增强位置估计的鲁棒性. 进一步设计指数加权状态平滑方法, 抑制状态估计的瞬时波动, 提高定位轨迹的连续性与稳定性. 仿真与实测结果表明, 所提方法的三维和高程均方根误差分别为7.59 m和6.06 m, 与对比方法中性能最优的双dog-leg增量估计(DDIE) 法相比分别降低约35.40% 和45.55%.Abstract: In constrained environments, global navigation satellite system (GNSS) may be limited by an insufficient number of visible satellites or degraded observation geometry. To address large disparities in observation quality and the tendency of nonlinear optimization to converge to undesirable local minima or even diverge in GNSS/cellular signals of opportunity fusion positioning, a method based on simulated annealing Levenberg-Marquardt (SA-LM) optimization is proposed. First, within a state estimation framework based on the iterated extended Kalman filter, a simulated annealing mechanism is incorporated into the Levenberg-Marquardt optimization process. Candidate steps that fail to satisfy the deterministic acceptance condition are accepted according to a probabilistic acceptance criterion to improve the reliability of iterative convergence under degraded observation conditions. A convergence analysis of the algorithm is also provided. Second, a dual-factor dynamic weighting mechanism driven by position dilution of precision and pseudorange residuals is developed to enhance the robustness of position estimation. Furthermore, an exponentially weighted state smoothing method is designed to suppress instantaneous fluctuations in the state estimate and improve the continuity and stability of the positioning trajectory. Simulation and field test results show that the proposed method achieves 3D and vertical RMSEs of 7.59 m and 6.06 m, respectively, representing reductions of approximately 35.40% and 45.55% compared with the double dog-leg incremental estimation (DDIE) method, which performs best among the comparison methods.
-
图 3 GNSS与蜂窝机会信号融合定位实验平台
Fig. 3 GNSS and cellular signals of opportunity fusion positioning experimental platform
图 5 两种典型北斗卫星构型的天空图
Fig. 5 Skyplots of two representative BeiDou satellite configurations
图 6 开阔环境下SA-LM与DDIE的定位轨迹对比
Fig. 6 Positioning trajectory comparison between SA-LM and DDIE in the open-sky environment
图 7 林荫道环境下不同蜂窝测距噪声条件的定位轨迹对比(北斗实测数据/蜂窝半仿真数据)
Fig. 7 Positioning trajectory comparison under different cellular ranging noise conditions in the tree-lined road environment (field test BDS data/semi-simulated cellular data)
图 8 加权融合模块消融实验定位误差对比
Fig. 8 Positioning error comparison of ablation experiments on the weighted fusion module
图 9 核心模块耦合消融实验定位误差对比
Fig. 9 Positioning error comparison of ablation experiments on core-module coupling
表 1 蜂窝基站信息
Table 1 Cellular base station information
小区标识 频段(MHz) 双工模式 坐标(纬度, 经度, 高度) 位置 135 1 835.6 FDD (112.854 8, 27.889 2, 116.3) 法学院 33 2 140.0 FDD (112.856 1, 27.891 1, 134.6) 第三教学楼 447 1 835.6 FDD (112.860 6, 27.884 8, 85.1) 琴湖宿舍 236 2 140.0 FDD (112.857 0, 27.885 8, 84.0) 南山阶梯教室 
下载: 导出CSV
表 2 测试场景与观测配置
Table 2 Test scenarios and observation configurations
场景 数据来源 配置项 参数取值 场景1
开阔环境北斗/蜂窝实测 PCI 135, 33, 477 E1卫星集合 C3, C35, C40 E2卫星集合 C1, C7, C26, C40, C59 场景2
林荫道环境北斗实测/
蜂窝半仿真可用卫星数 2 ~ 3颗 E3蜂窝测距噪声 $[\mu,\;\sigma]$=[5,5] m E4蜂窝测距噪声 $[\mu,\;\sigma]$=[10,8] m E5蜂窝测距噪声 $[\mu,\;\sigma]$=[20,10] m 重复次数 100次
下载: 导出CSV
表 3 两种北斗卫星/蜂窝基站退化构型下融合定位RMSE比较(m)
Table 3 RMSE comparison of fusion positioning under two degraded configurations of BDS satellites and cellular base stations (m)
卫星基站组合 方法 RMSE3D RMSEH RMSEV E1 LM[29] 74.79 12.35 73.76 WLS[17] 150670417.40 14149889.48 140901624.70 SDHA[15] 55.75 9.72 54.90 MOPSO[20] 75.04 12.53 73.98 DDIE[30] 11.75 3.77 11.13 SA-LM 7.59 4.57 6.06 E2 LM[29] 37.54 17.73 33.11 WLS[17] 34.80 16.23 30.81 SDHA[15] 23.85 10.44 21.46 MOPSO[20] 37.54 17.72 33.12 DDIE[30] 17.17 9.12 14.57 SA-LM 10.32 7.40 7.21
下载: 导出CSV
表 4 不同蜂窝测距噪声条件下融合定位RMSE比较(北斗实测数据/蜂窝半仿真数据, (m))
Table 4 RMSE comparison of fusion positioning under different cellular ranging noise conditions (field test BDS data/semi-simulated cellular data, (m))
方法 E3 E4 E5 RMSE3D RMSEH RMSEV RMSE3D RMSEH RMSEV RMSE3D RMSEH RMSEV LM[29] 43.74 10.96 42.34 53.56 15.73 51.20 71.04 21.52 67.71 WLS[17] — — — — — — — — — SDHA[15] 24.40 4.64 23.96 29.97 7.86 28.92 43.71 12.37 41.93 MOPSO[20] 3350.98 1055.43 3179.98 4538.81 1474.40 4292.84 4643.85 1513.76 4390.03 DDIE[30] 8.82 6.25 6.21 20.34 10.60 17.36 39.38 16.26 35.86 SA-LM 15.04 3.44 14.64 15.14 4.45 14.47 15.21 5.18 14.29
下载: 导出CSV
表 5 开阔环境下不同迭代优化算法的定位精度对比(m)
Table 5 Positioning accuracy comparison of different iterative optimization algorithms in the open-sky environment (m)
算法 E1 E2 RMSE3D RMSEH RMSEV RMSE3D RMSEH RMSEV LM-IEKF 8.32 4.57 6.97 10.49 7.42 7.45 BFGS-IEKF 41.72 8.60 40.82 70.76 23.38 68.93 DDIE-IEKF 7.90 4.59 6.43 10.50 7.40 7.48 SA-LM 7.59 4.57 6.06 10.32 7.40 7.21
下载: 导出CSV
表 6 林荫道环境下不同迭代优化算法的定位精度对比(北斗实测数据/蜂窝半仿真数据, (m))
Table 6 Positioning accuracy comparison of different iterative optimization algorithms in the tree-lined environment (field test BDS data/semi-simulated cellular data, (m))
算法 E3 E4 E5 RMSE3D RMSEH RMSEV RMSE3D RMSEH RMSEV RMSE3D RMSEH RMSEV LM-IEKF 15.49 3.47 15.09 15.61 4.58 14.91 15.67 5.37 14.71 BFGS-IEKF 33.75 7.15 32.83 39.98 9.26 38.78 38.61 9.21 37.35 DDIE-IEKF 15.50 3.47 15.10 15.61 4.58 14.92 15.68 5.38 14.71 SA-LM 15.04 3.46 14.64 15.14 4.45 14.47 15.21 5.18 14.29
下载: 导出CSV
表 7 不同迭代优化算法的计算效率比较
Table 7 Computational efficiency comparison of different iterative optimization algorithms
算法 单次迭代复杂度 平均迭代次数 平均单历元耗时(s) LM ${\rm{O}}(D_rD_x^2+D_x^3)$ 4.98 0.0164 BFGS ${\rm{O}}(D_rD_x+D_x^2)$ 18.27 0.0183 DDIE ${\rm{O}}(D_rD_x^2+D_x^3)$ 3.73 0.0021 SA-LM ${\rm{O}}(D_rD_x^2+D_x^3)$ 3.53 0.0018 注: Dx和Dr分别表示状态向量维数和残差向量维数.
下载: 导出CSV
表 8 加权融合模块消融实验配置
Table 8 Settings for ablation experiments on the weighted fusion module
配置 PDOP加权 伪距残差加权 状态平滑 基线 × × × PDOP加权 √ × × 双因子加权 √ √ × 双因子加权+平滑 √ √ √
下载: 导出CSV
表 9 所提方法优于各子组合的历元比例(%)
Table 9 Epoch ratio of the proposed method outperforming sub-combinations (%)
实验 < 1 Sub < 2 Subs < 3 Subs E1 100.00 93.89 20.56 E2 100.00 85.56 14.44 E5 100.00 88.30 29.82
下载: 导出CSV
表 10 核心模块耦合消融实验配置
Table 10 Settings for ablation experiments on core-module coupling
配置 SA-LM优化 双因子加权 状态平滑 基线 × × × 仅SA-LM √ × × 仅双因子加权 × √ × SA-LM + 双因子加权 √ √ × 完整方法 √ √ √
下载: 导出CSV
-
[1] Yu B, Du H, Wen G, et al. A vector-type finite-time output-constrained control algorithm and its application to a mobile robot. Robotic Intelligence and Automation, 2025, 45(3): 306−313 doi: 10.1108/RIA-04-2024-0087 [2] Zhu N, Marais J, Bétaille D, Berbineau M. GNSS position integrity in urban environments: a review of literature. IEEE Transactions on Intelligent Transportation Systems, 2018, 19(9): 2762−2778 doi: 10.1109/TITS.2017.2766768 [3] Ge Q, Gu F, Zhu Y, Zhu B, Zhang B, Wen G. Achieving autonomous landing of a UAV on a dynamic unmanned surface vessel: design, analysis and experiment. Robotic Intelligence and Automation, 2026, 46(3): 303−320 doi: 10.1108/RIA-06-2025-0156 [4] Zangenehnejad F, Gao Y. GNSS smartphones positioning: advances, challenges, opportunities, and future perspectives. Satellite Navigation, 2021, 2(1): 24 doi: 10.1186/s43020-021-00054-y [5] Liu Z D, Liu J C, Xu X W, Wu K S. DeepGPS: deep learning enhanced GPS positioning in urban canyons. IEEE Transactions on Mobile Computing, 2022, 23(1): 376−392 doi: 10.1109/tmc.2022.3208240 [6] Weng D J, Hou Z Y, Meng Y, Cai M M, Chan Y Y. Characterization and mitigation of urban GNSS multipath effects on smartphones. Measurement, 2023, 223: 113766 doi: 10.1016/j.measurement.2023.113766 [7] Liu S, Yao Z Q, Cao X M, Cai X W, Sheng M G. Enhancing navigation adaptability for ground vehicles: fine-grained urban environment recognition based on GNSS measurements. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 1−12 doi: 10.1109/tim.2024.3480202 [8] Chen Y Q, Shen H, Zhao D, Zhou J L, Wen G H. BITD: geometric-intensity fusion for robust LiDAR place recognition. Robotic Intelligence and Automation, 2025, 45(5-6): 637−645 doi: 10.1108/RIA-06-2025-0172 [9] Chen W, Zhang C, Peng Y, Yao Y, Cai M, Dong D. Enhancing GNSS positioning in urban canyon areas via a modified design matrix approach. IEEE Internet of Things Journal, 2024, 11(6): 10252−10265 doi: 10.1109/JIOT.2023.3326971 [10] Shamaei K, Kassas Z M. Receiver design and time of arrival estimation for opportunistic localization with 5G signals. IEEE Transactions on Wireless Communications, 2021, 20(7): 4716−4731 doi: 10.1109/TWC.2021.3061985 [11] Stock W, Schwarz R T, Hofmann C A, Knopp A. Survey on opportunistic PNT with signals from LEO communication satellites. IEEE Communications Surveys & Tutorials, 2025, 27(1): 77−107 doi: 10.1109/COMST.2024.3406990 [12] 董铭涛, 程建华, 赵琳, 刘萍. 惯性组合导航系统性能评估方法研究进展. 自动化学报, 2022, 48(10): 2361−2373 doi: 10.16383/j.aas.c210377Dong Ming-Tao, Cheng Jian-Hua, Zhao Lin, Liu Ping. Research progress on performance evaluation methods for integrated inertial navigation systems. Acta Automatica Sinica, 2022, 48(10): 2361−2373 doi: 10.16383/j.aas.c210377 [13] Kassas Z M, Khalife J, Abdallah A A, et al. I am not afraid of the GPS jammer: resilient navigation via signals of opportunity in GPS-denied environments. IEEE Aerospace and Electronic Systems Magazine, 2022, 37(7): 4−19 doi: 10.1109/MAES.2022.3154110 [14] 姚志强, 陈志鹏, 张文文, 曹学萌, 于澎. 移动射频测距辅助的卫星伪距误差抑制方法. 全球定位系统, 2026, 51(1): 49−57 doi: 10.12265/j.gnss.2025219Yao Zhi-Qiang, Chen Zhi-Peng, Zhang Wen-Wen, Cao Xue-Meng, Yu Peng. Mobile radio-frequency ranging-aided satellite pseudorange error suppression method. GNSS World of China, 2026, 51(1): 49−57 doi: 10.12265/j.gnss.2025219 [15] 龚福祥, 王庆, 张小国. 非视距环境下的GPS/TDOA最速下降混合定位算法. 仪器仪表学报, 2011, 32(7): 1500−1506 doi: 10.19650/j.cnki.cjsi.2011.07.009Gong Fu-Xiang, Wang Qing, Zhang Xiao-Guo. GPS/TDOA hybrid localization algorithm based on steepest descent in non-line-of-sight environments. Chinese Journal of Scientific Instrument, 2011, 32(7): 1500−1506 doi: 10.19650/j.cnki.cjsi.2011.07.009 [16] del Peral-Rosado J A, Saloranta J, Destino G, López-Salcedo J, Seco-Granados G. Methodology for simulating 5G and GNSS high-accuracy positioning. Sensors, 2018, 18(10): 3220 doi: 10.3390/s18103220 [17] Li F X, Tu R, Hong J, Zhang S X, Zhang P F, Lu X C. Combined positioning algorithm based on BeiDou navigation satellite system and raw 5G observations. Measurement, 2022, 190: 110763 doi: 10.1016/j.measurement.2022.110763 [18] De Angelis G, Baruffa G, Cacopardi S. GNSS/cellular hybrid positioning system for mobile users in urban scenarios. IEEE Transactions on Intelligent Transportation Systems, 2013, 14(1): 313−321 doi: 10.1109/TITS.2012.2215855 [19] Yin L, Ni Q, Deng Z L. A GNSS/5G integrated positioning methodology in D2D communication networks. IEEE Journal on Selected Areas in Communications, 2018, 36(2): 351−362 doi: 10.1109/JSAC.2018.2804223 [20] Lu J, Ding W Y, Wang W, Hu E W, Wu J F. A new hybrid positioning method by fusion of BDS and 5G signal using the Particle Swarm Method. Applied Sciences, 2022, 13(1): 366 doi: 10.3390/app13010366 [21] Li F X, Tu R, Geng F Y, Zhang R, Zhang P F, Lu X C. Combined BDS/5G seamless positioning scheme based on joint switching strategy of satellite elevation angle and CNR in complex urban environments. Measurement Science and Technology, 2024, 35(12): 126307 doi: 10.1088/1361-6501/ad7a1e [22] Li F X, Tu R, Hong J, Zhang P F, Wang J, Liu Y Y. High-precision GNSS/5G integrated positioning method and performance analysis for smartphones. Measurement Science and Technology, 2026, 37(1): 016303 doi: 10.1088/1361-6501/ae2644 [23] Sun C, Zhao H, Bai L, Cheong J W, Dempster A G, Feng W. GNSS-5G hybrid positioning based on TOA/AOA measurements. In: Proceedings of China Satellite Navigation Conference (CSNC). Singapore: Springer, 2020. 527-537 [24] Song W, Ding H, Zhang L, Chen L, Guo W, Lin W, Qi S. Performance verification of GNSS/5G tightly coupled fusion positioning in urban occluded environments with a smartphone. GPS Solutions, 2025, 29(1): 40 doi: 10.1007/s10291-024-01800-w [25] Zhou T, Chen R, Dong W, Ju Z, Jiao H, Liu J, et al. 5G NR robust tracking and positioning with GNSS assistance. Satellite Navigation, 2025, 6(1): 32 doi: 10.1186/s43020-025-00183-8 [26] Deng Z L, Zhang Z C, Ding Z K, Liu B X. Multi-source, fault-tolerant, and robust navigation method for tightly coupled GNSS/5G/IMU system. Sensors, 2025, 25(3): 965 doi: 10.3390/s25030965 [27] Bell B M, Cathey F W. The iterated Kalman filter update as a Gauss-Newton method. IEEE Transactions on Automatic Control, 1993, 38(2): 294−297 doi: 10.1109/9.250476 [28] Foy W H. Position-location solutions by Taylor-series estimation. IEEE Transactions on Aerospace and Electronic Systems, 1976, AES-12(2): 187−194 doi: 10.1109/TAES.1976.308294 [29] Skoglund M A, Hendeby G, Axehill D. Extended Kalman filter modifications based on an optimization viewpoint. In: Proceedings of 2015 18th International Conference on Information Fusion (Fusion). Washington, DC, USA: IEEE, 2015. 1856-1861 [30] Jin T, Zhang P, Chakwizira J, Wang Y C. Fusion positioning of GNSS-cellular signals of opportunity under low observability. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 1−18 doi: 10.1109/tim.2024.3485434 [31] Goffin S, Barrau A, Bonnabel S, Brüls O, Sacré P. Iterated invariant extended Kalman filter (IterIEKF). IEEE Transactions on Automatic Control, 2026, 71: 3380−3387 doi: 10.1109/TAC.2025.3637661 [32] Bai L, Sun C, Dempster A G, Zhao H B, Cheong J W, Feng W Q. GNSS-5G hybrid positioning based on multi-rate measurements fusion and proactive measurement uncertainty prediction. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1−15 doi: 10.1109/tim.2022.3154821 [33] Jiao H J, Tao X X, Chen L, Zhou X, Ju Z H. GNSS/5G joint position based on weighted robust iterative Kalman filter. Remote Sensing, 2024, 16(6): 1009 doi: 10.3390/rs16061009 [34] Langley R B. Dilution of precision. GPS world, 1999, 10(5): 52−59 [35] Zekavat R, Buehrer R M. Handbook of Position Location: Theory, Practice, and Advances. Hoboken, NJ: Wiley-IEEE Press, 2019. 22-40 [36] Wang B, Song B, Wang T, Deng Z H, Fu M Y. A BDS/5G combined positioning method based on adaptive optimal selection-robust hybrid adaptive Kalman filter algorithm. IEEE Internet of Things Journal, 2024, 11(12): 22376−22384 doi: 10.1109/JIOT.2024.3381695 [37] Cao S, Chen X, Zhang X, Chen X. Combined weighted method for TDOA-based localization. IEEE Transactions on Instrumentation and Measurement, 2019, 69(5): 1962−1971 [38] Campolo F, Blaga A, Rea M, Lozano A, Costa-Pérez X. 5GNSS: fusion of 5G-NR and GNSS localization for enhanced positioning accuracy and reliability. IEEE Transactions on Vehicular Technology, 2024, 73(9): 13558−13568 doi: 10.1109/TVT.2024.3396991 [39] Huang Z J, Xu Q S, Sun M H, Zhu X Z. A Kalman filter-based localization calibration method optimized by reinforcement learning and information matrix fusion. Entropy, 2025, 27(8): 821 doi: 10.3390/e27080821 [40] Ge Q B, Lin K, Zhao Z Y. Credibility-based multi-sensor fusion for non-Gaussian conversion error mitigation. Information Fusion, 2025, 115: 102704 doi: 10.1016/j.inffus.2024.102704 [41] Alghisi M, Biagi L. Positioning with GNSS and 5G: analysis of geometric accuracy in urban scenarios. Sensors, 2023, 23(4): 2181 doi: 10.3390/s23042181 [42] Li F X, Tu R, Zeng L C, Zhang S X, Liu M Y, Lu X C. Integrated positioning with double-differenced 5G and undifferenced/double-differenced GPS. Measurement, 2023, 218: 113114 doi: 10.1016/j.measurement.2023.113114 [43] 刘浩洋, 户将, 李勇锋, 文再文. 最优化计算方法. 北京: 高等教育出版社, 2022. 176-184Liu Hao-Yang, Hu Jiang, Li Yong-Feng, Wen Zai-Wen. Optimization Calculation Methods. Beijing: Higher Education Press, 2022. 176-184 [44] Golub G H, Van Loan C F. Matrix Computations (4th edition). Baltimore: Johns Hopkins University Press, 2013. 87-90 -
下载:
计量
- 文章访问数: 1
- HTML全文浏览量: 0
- 被引次数: 0
