基于自适应变分贝叶斯UKF的高动态三维目标跟踪方法

High-dynamic Three-dimensional Target Tracking Method Based on Adaptive Variational Bayesian UKF

摘要

在高动态条件下,目标高速移动或状态快速变化导致系统模型的不确定性与量测噪声的复杂性显著增加,传统基于线性或固定噪声假设的滤波方法难以保持稳定的估计精度。针对高动态炮弹三维跟踪任务中系统模型强非线性、噪声统计特性未知及时变等问题,提出一种基于自适应变分贝叶斯的无迹卡尔曼滤波算法(unscented Kalman filter,UKF)。该方法在UKF的框架中引入逆Wishart先验分布,结合变分贝叶斯推理(variational Bayesian,VB),实现对系统噪声与量测噪声协方差的在线估计,从而自适应地调整滤波增益。为进一步改善传统VB方法收敛性差、计算复杂度高的问题,设计了基于滤波收敛准则的自适应迭代策略,并通过对VB估计项的非线性修正,确保状态协方差与噪声统计特性的一致性。分别在高斯噪声、时变噪声及野值污染3种典型环境下开展仿真实验,并与UKF、Sage-Husa UKF及VB-UKF等算法进行对比。结果表明,所提方法在多种噪声环境下均能实现稳定收敛,在野值污染条件下位置估计精度相比最优对比算法提升约15.3%,并在时变噪声与突发干扰下仍保持较高的鲁棒性与适应性。

Abstract

Under highly dynamic conditions involving high-speed moving targets or rapidly changing system states, the uncertainty of the system model and the complexity of measurement noise increase significantly. Conventional Kalman-type filters based on linear or fixed-noise assumptions often fail to achieve stable estimation accuracy. To address the issues of model nonlinearity and noise uncertainty in three-dimensional tracking of high-dynamic artillery shells, based on an adaptive variational Bayesian (VB), unscented Kalman filter (UKF) with an inverse-Wishart prior distribution was proposed in this paper. The proposed method integrates the VB inference framework with the UKF through a nonlinear correction mechanism, and establishes an adaptive iterative strategy guided by the convergence criterion of filtering. This design enables the online estimation of both process and measurement noise covariances, thus improving robustness against non-stationary and uncertain noise. Furthermore, a modified covariance propagation was introduced to ensure consistency between state prediction and the dynamically updated noise statistics. Simulation experiments were conducted under three-typical noise environments: Gaussian, time-varying, and outlier-contaminated conditions with comparisons against conventional UKF, Sage-Husa UKF, and VB-UKF algorithms. The results demonstrate that the proposed method achieves stable convergence and superior tracking accuracy. Under outlier contamination, it improves position estimation accuracy by approximately 15.3% over the best-performing benchmark algorithm, maintaining strong adaptability and robustness under time-varying and abrupt disturbances noise environments.

关键词

高动态炮弹 / 目标跟踪 / 自适应滤波 / 变分贝叶斯

Key words

high-dynamic projectile / target tracking / adaptive filter / variational bayesian

参考文献

[1] 刘增力, 张文, 曹奇宏, 等. 面向三维目标跟踪的状态转换卡尔曼滤波方法[J]. 兵工学报, 2024, 45(11): 3998-4010.

[2] LIU Zengli, ZHANG Wen, CAO Qihong, et al. Converted state equation Kalman filter for three-dimensional target tracking[J]. Acta Armamentarii, 2024, 45(11): 3998-4010. (in Chinese)

[3] DUNIK J, BISWAS S, DEMPSTER A, et al. State estimation methods: overview and application in navigation[J]. IEEE Aerospace and Electronic Systems Magazine, 2020, 35(12): 16-31.

[4] ZHANG W, MA X, WANG S, et al. Adaptive unscented Kalman filter based on sequential state difference for spacecraft autonomous navigation during the approach phase[J]. Measurement, 2024, 243: 116330.

[5] 陈之曦, 王健. 纯方位目标跟踪中相对距离的估计方法[J]. 弹道学报, 2025, 37(1): 54-59.

[6] CHEN Zhixi, WANG Jian. Estimation method for relative distance in bearings-only target tracking[J]. Journal of Ballistics, 2025, 37(1): 54-59. (in Chinese)

[7] YU X, LI J. Adaptive Kalman filtering for recursive both additive noise and multiplicative noise[J]. IEEE Transactions on Aerospace and Electronic Systems, 2021, 58(3): 1634-1649.

[8] MENG Y, GAO S, ZHONG Y, et al. Covariance matching based adaptive unscented Kalman filter for direct filtering in INS/GNSS integration[J]. Acta Astronautica, 2016, 120: 171-181.

[9] HU G, GAO B, ZHONG Y, et al. Unscented Kalman filter with process noise covariance estimation for vehicular INS/GPS integration system[J]. Information Fusion, 2020, 64: 194-204.

[10] GAO Z, MU D, GAO S, et al. Adaptive unscented Kalman filter based on maximum posterior and random weighting[J]. Aerospace Science and Technology, 2017, 71: 12-24.

[11] GAO B, GAO S, HU G, et al. Maximum likelihood principle and moving horizon estimation based adaptive unscented Kalman filter[J]. Aerospace Science and Technology, 2018, 73: 184-196.

[12] LIN X, LIANG X, LI W, et al. Adaptive robust least-squares smoothing algorithm[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-18.

[13] ZHU H, ZHANG G, LI Y, et al. An adaptive Kalman filter with inaccurate noise covariances in the presence of outliers[J]. IEEE Transactions on Automatic Control, 2021, 67(1): 374-381.

[14] GUI M, YANG H, NING X, et al. Variational Bayesian implicit unscented Kalman filter for celestial navigation using time delay measurement[J]. Advances in Space Research, 2023, 71(1): 756-767.

[15] ZHANG Y, LI M, ZHANG Y, et al. An enhanced adaptive unscented Kalman filter for vehicle state estimation[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-12.

[16] GAO B, LI W, HU G, et al. Mahalanobis distance-based fading cubature Kalman filter with augmented mechanism for hypersonic vehicle INS/CNS autonomous integration[J]. Chinese Journal of Aeronautics, 2022, 35(5): 114-128.

[17] HU G, XU L, YANG Z, et al. Indirect fuzzy robust cubature Kalman filter with normalized input parameters[J]. IEEE Transactions on Aerospace and Electronic Systems, 2024, 60(5): 5880-5890.

[18] GAO B, HU G, ZHANG L, et al. Cubature Kalman filter with closed-loop covariance feedback control for integrated INS/GNSS navigation[J]. Chinese Journal of Aeronautics, 2023, 36(5): 363-376.

[19] HUANG Y, ZHANG Y, WU Z, et al. A novel adaptive Kalman filter with inaccurate process and measurement noise covariance matrices[J]. IEEE Transactions on Automatic Control, 2017, 63(2): 594-601.

[20] FENG K, WANG J, WANG X, et al. Adaptive state estimation and filtering for dynamic positioning ships under time-varying environmental disturbances[J]. Ocean Engineering, 2024, 303: 117798.

[21] DUTRA D A. Parameterizations for large-scale variational system identification using unconstrained optimization[J]. Automatica, 2025, 173: 112086.

[22] ZHANG X, CHEN Z, YU X, et al. A novel two-stage DOA estimation of sound sources based on hierarchical sparse Bayesian inference[J]. Applied Acoustics, 2025, 231: 110514.

[23] ZOU H, WU S, XUE Q, et al. A novel Gaussian-Student's t-Skew mixture distribution based Kalman filter[J]. Signal Processing, 2025, 230: 109787.

[24] PAN C, GAO J, LI Z, et al. Multiple fading factors-based strong tracking variational Bayesian adaptive Kalman filter[J]. Measurement, 2021, 176: 109139.

[25] LIU X, LIU X, YANG Y, et al. Variational Bayesian-based robust cubature Kalman filter with application on SINS/GPS integrated navigation system[J]. IEEE Sensors Journal, 2021, 22(1): 489-500.

[26] 贲思铭, 马国梁, 许立松, 等. 一种基于SR-CDKF滤波的弹道目标雷达跟踪算法研究[J]. 弹道学报, 2025, 37(3): 25-32.

[27] BEN Siming, MA Guoliang, XU Lisong, et al. Research on a ballistic radar target tracking algorithm based on SR-CDKF filter[J]. Journal of Ballistics, 2025, 37(3): 25-32. (in Chinese)

[28] ZHANG T, ZHAO S, LUAN X, et al. Bayesian inference for state-space models with student-t mixture distributions[J]. IEEE Transactions on Cybernetics, 2022, 53(7): 4435-4445.

基金资助

国家自然科学基金青年基金(62403243)

中国科协青年人才托举(2023QNRC001)

国家自然科学基金基础科学中心项目(62388101)

中央高校基本科研业务费专项资金资助(30924010931)