康英杰,陈启丽,郑智民,陈雯柏

• 1:北京信息科技大学自动化学院
• 2:中国移动创新研究院

摘要(Abstract)：

惯性姿态估计（inertial orientation estimation, IOE）模型是运动感知与导航中的关键基础，其精度决定了后续任务链的安全性与可靠性。针对现有模型存在参数依赖或产生非物理解导致精度与稳定性不足的问题，提出一种基于物理信息神经网络（physics-informed neural network, PINN）的IOE模型。该模型以并行架构为载体嵌入物理约束，从变量角度映射高维特征表示，拟合物理量间的内在关联并解析物理算子；在特征交互过程中施加筛选机制，进一步抑制异步采样点间的冗余信息，形成了无参数依赖、具备物理一致性的全新模型。对比8种现有模型，该模型整体姿态误差仅为2.720 3°，相对误差降幅最低可达3.7%。针对数据驱动模型对条件变化的敏感性，进一步验证该模型在不同运动方向及安装姿态下的性能。结果显示，模型整体姿态误差均值为4.285 3°，性能退化不超过2°，对工况变化表现出较好的适用性与稳定性。

关键词(KeyWords)： 惯性姿态估计(inertial orientation estimation, IOE);物理信息神经网络(physicsinformed neural network, PINN);物理一致性;传感器融合

Abstract:

Keywords:

基金项目(Foundation): 北京市自然科学基金-小米创新联合基金重点研究专题(L233006);; 国家自然科学基金重大研究计划项目(92267110);国家自然科学基金项目(62276028,62103056,62576044);; 北京信息科技大学勤信拔尖人才项目(QXTCP B202403)

作者(Author): 康英杰,陈启丽,郑智民,陈雯柏

DOI: 10.16508/j.cnki.11-5866/n.2026.03.009

参考文献(References)：

• [1]ALLIONE F,GAMBA J D,GKIKAKIS A E,et al. Effects of repetitive low-acceleration impacts on attitude estimation with micro-electromechanical inertial measurement units[J].Frontiers in Robotics and AI,2023,10:1211531.
• [2]ENGELSMAN D, KLEIN I. Information-aided inertial navigation:a review[J]. IEEE Transactions on Instrumentation and Measurement,2023,72:8507418.
• [3]NAZARAHARI M,ROUHANI H. 40 years of sensor fusion for orientation tracking via magnetic and inertial measurement units:methods,lessons learned,and future challenges[J].Information Fusion,2021,68:67-84.
• [4]蒋维，章卫国，史静平，等.一种基于MEMS传感器的全局渐进稳定的姿态估计算法[J].西北工业大学学报，2020,38(3):550-557.JIANG W,ZHANG W G,SHI J P,et al. Globally asymptotic stable attitude estimation with application to MEMS sensors[J].Journal of Northwestern Polytechnical University,2020,38(3):550-557.(in Chinese)
• [5]JUSTA J，??M??DL V,HAM????EK A. Fast AHRS filter for accelerometer,magnetometer,and gyroscope combination with separated sensor corrections[J]. Sensors,2020,20(14):3824.
• [6]HALL T,RASHEED K. A survey of machine learning methods for time series prediction[J]. Applied Sciences,2025,15(11):5957.
• [7]WANG S F,TENG Y J,PERDIKARIS P. Understanding and mitigating gradient flow pathologies in physics-informed neural networks[J]. SIAM Journal on Scientific Computing,2021,43(5):A3055-A3081.
• [8]MADGWICK S O H. An efficient orientation filter for inertial and inertial/magnetic sensor arrays[R]. Bristol:University of Bristol,2010.
• [9]MAHONY R, HAMEL T, PFLIMLIN J M. Nonlinear complementary filters on the special orthogonal group[J]. IEEE Transactions on Automatic Control,2008,53(5):1203-1218.
• [10]胡启国，王磊，马鉴望，等.基于IQPSO-EKF的多传感器融合姿态测量方法研究[J].机电工程，2024,41(2):353-363.HU Q G,WANG L,MA J W,et al. Attitude measurement of multi-sensor fusion based on IQPSO-EKF[J]. Journal of Mechanical&Electrical Engineering,2024,41(2):353-363.(in Chinese)
• [11]LAIDIG D,SEEL T. VQF:highly accurate IMU orientation estimation with bias estimation and magnetic disturbance rejection[J]. Information Fusion,2023,91:187-204.
• [12]NARKHEDE P,WALAMBE R,PODDAR S,et al. Incremental learning of LSTM framework for sensor fusion in attitude estimation[J]. PeerJ Computer Science,2021,7:e662.
• [13]WEBER D,G??HMANN C,SEEL T. RIANN:a robust neural network outperforms attitude estimation filters[J]. AI,2021,2(3):444-463.
• [14]BROTCHIE J,SHAO W,LI W C,et al. Leveraging selfattention mechanism for attitude estimation in smartphones[J].Sensors,2022,22(22):9011.
• [15]GREYDANUS S,DZAMBA M,YOSINSKIJ. Hamiltonian neural networks[C]//33rd Conference on Neural Information Processing Systems(NeurIPS). La Jolla, CA, USA:NIPS,2019,32:15379-15389.
• [16]LI Z Y,HUANG D Z,LIU B,et al. Fourier neural operator with learned deformations for PDEs on general geometries[J]. Journal of Machine Learning Research,2023,24(1):18593-18618.
• [17]VASWANI A,SHAZEER N,PARMAR N,et al. Attention is all you need[C]//31st Annual Conference on Neural Information Processing Systems(NIPS). La Jolla, CA,USA:NIPS,2017,30:1-15.
• [18]VALENTI R G,DRYANOVSKI I,XIAO J Z. Keeping a good attitude:a quaternion-based orientation filter for IMUs and MARGs[J]. Sensors,2015,15(8):19302-19330.
• [19]郭津延，郑红，杜佳宇，等.基于双尺度自适应令牌注意力的交通流量预测[J].华东理工大学学报（自然科学版），2025,51(6):817-826.GUO J Y,ZHENG H,DU J Y,et al. Traffic flow prediction based on dual-scale adaptive token attention[J]. Journal of East China University of Science and Technology(Natural Science Edition),2025,51(6):817-826.(in Chinese)
• [20]LAIDIG D,CARUSO M,CEREATTI A,et al. BROAD:a benchmark for robust inertial orientation estimation[J]. Data,2021,6(7):72.
• [21]GU H Y,JIN C C,YUAN H Y,et al. Design and implementation of attitude and heading reference system with extended Kalman filter based on MEMS multi-sensor fusion[J]. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems,2021,29(Supp1):157-180.
• [22]CHIELLA A C B,TEIXEIRA B O S,PEREIRA G A S.Quaternion-based robust attitude estimation using an adaptive unscented Kalman filter[J]. Sensors,2019,19(10):2372.
• [23]SIMOULIN A,PARK N,LIU X Y,et al. Memory-efficient finetuning of transformers via token selection[C]//Proceedings of the2024 Conference on Empirical Methods in Natural Language Processing. Stroudsburg, PA, USA:ACL, 2024:21565-21580.