在机动目标跟踪领域，可靠和及时的目标机动检测是目标运动状态估计的关键。因此，机动检测方法一直被视为目标跟踪的重要前提，目前主要存在以下两方面难题：（1）机动目标运动模型不确定，属于典型的多分布检测（Multiple Distribution Detecton，MDD）问题；（2）目标机动检测最快决策，属于典型的变点检测（Change Point Detection）问题。基于决策的机动目标跟踪对检测速率要求非常高，最小化检测延迟往往可以提高目标跟踪精度。本文融合多模型假设检验（Multiple Model Hypothesis Testing，MMHT）与Shiryayev序贯概率比检验（Shiryayev Sequential Probability Ratio Test，SSPRT）算法框架实现目标机动检测，主要贡献如下：

（1）针对现有MMHT方法中模型集似然（Model-Set Likelihood，MSL）不一定具有代表性的问题，本文提出了一种使用代表性模型（Representative Model，RM）来检测可能具有多个分布的未知事件的MMHT方法，称为基于代表性模型的多模型假设检验（MMHT-RM）方法。首先，在特定标准下寻找模型集覆盖区域中最可能接近真实的模型；其次，将寻找到的模型增广至原始模型集；最后，通过运行多模型（Multiple-Model，MM）算法，选择具有最大概率的模型作为模型覆盖区域的RM。它们解决了具有相关观测的复合、多元、相交和错误指定假设集的MMHT等难题。

（2）本文设计了两种适用于不同场景的模型自适应方法提高模型集的覆盖能力，分别为基于RM的期望模型增强（Expected-Mode Augmentation，EMA）和最佳模型增强（Best Model Augmentation，BMA）。提出的RM-EMA使用固定模型的最小均方误差估计来增强基本模型集，适用于处理具有相同物理意义（可加性）参数的模型集。提出的RM-BMA可以处理具有不同结构或参数的候选模型，并且是在模式空间中找到在Kullback-Leiber（K-L）准则下最接近真实的模型。

（3）针对目标机动的变点检测最优决策问题，本文提出一种新的多模型框架下的序贯检测方法，采用著名SSPRT来作为机动检测器，检测器中涉及的似然函数均由MM方法输出的RM的似然近似。基于自适应多模型中不同参数和不同结构模型集的场景，分别提出基于RM-EMA和RM-BMA的Shiryayev多模型假设检验方法：EMA-MMSSPRT和BMA-MMSSPRT。

本文通过设计运动模型的目标机动检测仿真实验进行验证，仿真结果证明了所提出的算法的优效性。与目前流行的MMSSPR、MMCUSUM等多模型机动检测算法相比，提出的检测方法不仅提高目标机动的检测精度，而且可以降低平均检测延迟。

In the field of maneuvering target tracking, reliable and timely target maneuvering detection is the key to estimating target motion states. Therefore, the maneuvering detection method has always been regarded as an important prerequisite for target tracking, and currently there are mainly two challenges: (1) the uncertainty of the maneuvering target motion model, which belongs to the typical Multiple Distribution Detection (MDD) problem; (2) The fastest decision for target maneuvering detection belongs to the typical Change Point Detection problem. Decision based maneuvering target tracking requires very high detection rate, and minimizing detection delay can often improve target tracking accuracy. This article integrates the Multiple Model Hypothesis Testing (MMHT) and Shiryaev Sequential Probability Ratio Test (SSPRT) algorithm frameworks to achieve target maneuver detection. The main contributions are as follows.

(1) In response to the problem that Model Set Likelihood (MSL) may not be representative in existing MMHT methods, this paper proposes an MMHT method that uses Representative Model (RM) to detect unknown events that may have multiple distributions. This method is called Representative Model Based Multiple Model Hypothesis Testing (MMHT-RM). Firstly, search for the model that is closest to the real model in the coverage area of the model set under specific standards; Secondly, expand the found model to the original model set; Finally, by running the Multiple Model (MM) algorithm, select the model with the highest probability as the RM for the model coverage area. They solve the challenges of MMHT with composite, multivariate, intersecting, and incorrectly specified hypothesis sets with relevant observations.

(2) This paper proposes two model adaptation methods suitable for different scenarios to improve the coverage ability of the model set, namely Expected Model Augmentation (EMA) and Best Model Augmentation (BMA). RM-EMA uses the minimum mean square error estimation of fixed models to enhance the basic model set, suitable for processing model sets with parameters of the same physical meaning (additivity). RM-BMA can handle candidate models with different structures or parameters, and it is the model found in the pattern space that is closest to the real model under the Kullback Leiber (K-L) criterion.

(3) This paper proposes a new sequential detection method under a multi model framework for the optimal decision problem of target maneuvering change point detection. The famous SSPRT is used as the maneuvering detector, and the likelihood functions involved in the detector are all approximated by the likelihood of the RM output by the MM method. Based on scenarios with different parameters and structural model sets in adaptive multiple models, we propose Shiryayev multiple model hypothesis testing methods based on RM-EMA and RM-BMA, respectively: EMA-MMSSPRT and BMA-MMSSPRT.

This paper verifies the effectiveness of the proposed algorithm by designing a target maneuver detection simulation experiment with a motion model, and the simulation results demonstrate its effectiveness. Compared with current popular multiple model maneuver detection algorithms such as MMSSPR and MMCUSUM, the proposed methods not only improve the detection accuracy of target maneuver, but also reduce the average detection delay.

[1] 韩崇昭, 朱洪艳, 段战胜, 等. 多源信息融合[M]. 北京: 清华大学出版社, 2006.

[2] 周弘仁, 敬忠良, 王培德. 机动目标跟踪[M]. 北京: 国防工业出版社, 1991.

[3] D. F. Bizup, D. B. Brown. Maneuver detection using the range rate measurement[J]. IEEE Transactions on Aerospace and Electronic Systems, 2004, 40(1): 330-336．

[4] M. Basseville, I. Nikiforov. Detection of abrupt changes: theory and application[M]. Englewood Cliffs, NJ: Prentice Hall, 1993．

[5] E. S. Page. A test for a change in a parameter occurring at an unknown point[J]. Biometrika, 1955, 42(4): 523-527.

[6] X. R. Li, P. J. Vesselin. A survey of maneuvering target tracking. Part Ⅴ: multiple-model methods[J]. IEEE Transactions on Aerospace and Electronic Systems, 2005, 41(4): 1255-1321.

[7] D. Gaglione, G. Soldi, F. Meyer, et al. Bayesian information fusion and multitarget tracking for maritime situational awareness[J]. IET Radar, Sonar & Navigation, 2020, 14(12): 1845-1857.

[8] 赵玉杰, 杨晨, 宋琛. 空基高超声速导弹防御系统关键技术研究[J]. 战术导弹技术, 2020, 10(4): 64-70.

[9] 朱自谦. 机载雷达多目标跟踪技术研究[D]. 南京: 南京航空航天大学, 2010.

[10] S. Li, Y. Cheng, H. Wang, et al. Distributed multisensor multitarget tracking algorithm with time-offset registration[J]. Journal of Northwestern Polytechnical University, 2020, 38(4): 797-805.

[11] X. R. Li, Y. Bar-Shalom. Design of an interacting multiple model algorithm for air traffic control tracking[J]. IEEE Transactions on Control Systems Technology, 1993, 1(3): 186-194.

[12] 刘妹琴, 兰剑. 目标跟踪前沿理论与应用[M]. 科学出版社, 2015.

[13] P. L. Bogler. Tracking a maneuvering target using input estimation[J]. IEEE Transactions on Aerospace and Electronic Systems, 1987, 23(3): 298-310.

[14] H. Lee, M. J. Tahk. Generalized input-estimation technique for tracking maneuvering targets[J]. IEEE Transactions on Aerospace and Electronic Systems, 1999, 35(4): 1388-1402.

[15] Y. T. Chan, F. Couture. Maneuver detection and track correction by input estimation[J]. IEE Proceedings-F, 1993, 140(1): 21-28.

[16] Y. Bar-Shalom, K. Birmiwal. Variable dimension filter for maneuvering target tracking[J]. IEEE Transactions on Aerospace and Electronic Systems, 1982, 18(5): 621-629.

[17] X. R. Li. Multiple-model estimation with variable structure. Part II: model-set adaptation[J]. IEEE Transactions on Automatic Control, 2000, AC-45(11): 2047-2060.

[18] X. R. Li, V. P. Jilkov. A survey of maneuvering target tracking. Part IV: decision-based methods[C]//Proceedings of 2002 SPIE Conference on Signal and Data Processing of Small Targets, 2002, 4728.

[19] X. R. Li, V. P. Jilkov. Survey of maneuvering target tracking. Part I: dynamic models[J]. IEEE Transactions on Aerospace and Electronic Systems, 2003, 39(4): 1333-1364.

[20] F. Gustafsso. The marginalized likelihood test for detecting abrupt changes[J]. IEEE Transactions on Automatic Control, 1996, 41: 66-78.

[21] A. S. Willsky, H. L. Jones．A generalized likelihood ratio approach to the estimation in linear systems subject to abrupt changes[C]∥Proceedings of IEEE Conference on Decision and Control, 1974．

[22] A. S. Willsky, H. L. Jones．A generalized likelihood ratio approach to the detection and estimation of jumps in linear systems[J]. IEEE Transactions on Automatic Control, 1976, 21: 108-112.

[23] J. Korn, S. W. Gully, A. S. Willsky．Application of the generalized likelihood ratio algorithm to maneuver detection and estimation[C]∥Proceedings of the American Control Conference, 1982.

[24] K. Schnepper．A comparison of GLR and multiple model filters for a target tracking problem[C]∥Proceedings of the 25th Conference on Decision and Control, 1986, 666-670．

[25] K. Konopko, D. Janczak．Maneuver detection algorithm based on probability density function estimation with use of RBF and HRBF neural network[C]∥Proceedings of SPIE on Photonics Applications in Astronomy, Communications, Industry, and High-Energy, 2004, 5484: 625-629.

[26] X. R. Li, L. Wang. Fault detection using sequential probability ratio test[C]//Proceedings Power Engineering Society 1999 Winter Meeting, 1999, 2: 938-943.

[27] R. Chandramouli, N. D. Memon. On sequential watermark detection[J]. IEEE Transactions on Signal Processing, 2003, 51(4): 1034-1044.

[28] A. G. Tartakovsky, I. A. Ivanova. Approximations in sequential rules for discrimination of composite hypotheses and their precision in the problem of signal detection from post-detector data[J]. Problems of Information Transmission, 1992, 28(1): 55-66.

[29] C. Leang, D. Johnson, On the asymptotic of M-hypothesis Bayesian detection[J]. IEEE Transactions on Information Theory, 1997, 43(1): 280-282.

[30] 范红旗, 翟庆林, 王胜, 等. 锐截止中频采样滤波器的设计与实现[J]. 电路与系统学报, 2009, 14(5): 31-36.

[31] 濮晓龙. 关于累积和（CUSUM）检验的改进[J]. 应用数学学报, 2003, 26(2): 225-241.

[32] Chris J. OliverbaN. A comparison of segmentation methods with standard CFAR for point target detection[J]. SPIE, 1998, 3497: 76-87.

[33] G. Emanuele, L. Marco. Sequential detection of Markov targets with trajectory estimation[J]. IEEE Transactions on Information Theory, 2008, 54(9): 4144-4154.

[34] E. S. Page. A test for a change in a parameter occurring at an unknown point[J]. Biometrika, 1995, 42(4): 523-527.

[35] S. W. Roberts. Control chart tests based on geometric moving average[J]. Technometrics, 1959, 1: 239-250.

[36] T. L. Lai, Sequential multiple hypothesis testing and efficient fault detection-isolation in stochastic systems[J]. IEEE Transactions on Information Theory, 2000, 46(2): 595-608.

[37] G. V. Moustakides. Optimal stopping times for detecting changes in distributions[J]. The Annals of Statistic, 1977, 14(4): 1379-1387.

[38] A. N. Shiryayev, Optimal Stopping Rules[M]. New York: Springer-Verlag, 1977.

[39] S. Li, H. Lei, L. Shao, et al. Multiple model tracking for hypersonic gliding vehicles with aerodynamic modeling and analysis[J]. IEEE Access, 2019, 7: 28011-28018.

[40] H. A. P. Blom. An efficient filter for abruptly changing systems[C]//The 23rd IEEE Conference on Decision and Control, 1984, 656-658.

[41] S. zhao, C. K. Ahn, P. Shi, et al. Bayesian state estimation for Markovian jump systems: employing recursive steps and pseudocodes[J]. IEEE Systems, Man, and Cybernetics Magazine, 2019, 5(2): 27-36.

[42] G. Yao, R. Saltus, A. Dani. Image moment-based extended object tracking for complex motions[J]. IEEE Sensors Journal, 2020, 20(12)12: 6560-6572.

[43] W. Youn, H. Myung. Robust interacting multiple model with modeling uncertainties for maneuvering target tracking[J]. IEEE Access, 2019, 7: 65427-65443.

[44] Blom. H. A. P. An efficient filter for abruptly changing systems[C]//Proceedings of 23rd Conference on Decision and Control, 1984, 656-658.

[45] M. Xiang, B. Scalzo Dees, D. P. Mandic. Multiple-model adaptive estimation for 3-D and 4-D signals: a widely linear quaternion approach[J]. IEEE Transactions on Neural Networks and Learning Systems,2019, 30(1): 72-84, 2019.

[46] L. Chang, B. Hu, A. Li, F. Qin. Transformed unscented Kalman filter[J]. IEEE Transactions on Automatic Control, 2013, 58(1): 252-257.

[47] J. Ru, V. P. Jilkov, X. R. Li, A. Bashi. Detection of target maneuver onset[J]. IEEE Transactions on Aerospace and Electronic Systems, 2009, 45(2): 536-554.

[48] Y. Liu, X. R. Li. Sequential multiple-model detection of target maneuver termination[C]//Proceedings 14th International Conference Information Fusion, 2011, 873-880.

[49] B. Liu, J. Lan, X. R. Li. A 2-SPRT based approach to multiple-model hypothesis testing for multi-distribution detection[J]. IEEE Transactions on Signal Processing, 2016, 64(12): 3221-3236.

[50] B. Liu, H. Yang, X. Zhai. Maximum correntropy AMM filter for linear systems with non-Gaussian noise[C]//2023 42nd Chinese Control Conference (CCC), 2023, 3145-3151.

[51] T. Oskiper, H. Poor. Online activity detection in a multiuser environment using the matrix CUSUM algorithm[J]. IEEE Transactions on Information Theory, 2002, 48(2): 477-493.

[52] E. S. Page. Continuous inspection schemes[J]. Biometrika, 41: 100-115.

[53] V. Rossi, J. P. Vila. Approche non paramétrique du filtrage de système non linéaireà temps discret età paramètres inconnus[J]. C. R.Acad. Sci. Paris, Ser. I, 2005, 340: 759-764.

[54] D. P. Malladi, J. L. Speyer. A generalized Shiryaev sequential probability ratio test for change detection and isolation[J]. IEEE Transactions on Automatic Control, 1999, 44(8): 1522-1534.

[55] I. Nikiforov. Optimal sequential change detection and isolation[C]// Proceedings 15th IFAC Word Congress, 2002.

[56] 赵玉杰, 杨晨, 宋琛. 空基高超声速导弹防御系统关键技术研究[J]. 战术导弹技术, 2020, 10(4): 64-70.

[57] G. Wang, R. Xue, J. Wang. A distributed maximum correntropy Kalman filter[J]. Signal Processing, 2019, 160(1): 247-251.

[58] X. Fan, G. Wang, J. Han, Y. Wang. Interacting multiple model based on maximum correntropy Kalman filter[J]. IEEE Transactions on Circuits and Systems II: Express Briefs, 2021, 68(8): 3017-3021.

[59] P. S. Maybeck, R. D. Stevens. Reconfigurable flight control viamultiple model adaptive control method[J]. IEEE Transactions on Aerospace and Electronic Systems, 1991, 27(3)3: 470-480.

[60] P. S. Maybeck. Application of multiple model adaptive algorithms to reconfigurable flight control[J]. Control and Dynamic Systems, 1992, 52: 291-320.

[61] C. Leang, D. Johnson. On the asymptotic of M-hypothesi bayesian detection[J]. IEEE Transactions on Information Theory, 1997, 43(1): 280-282.

[62] S. Zacks. The theory of statistical inference[M]. New York, USA:Wiley, 1971.

[63] C. D. Richmond, L. L. Horowitz. Parameter bounds on estimation accuracy under model misspecification[J]. IEEE Transactions on Signal Processing, 2015, 63(9): 2263-2278.

[64] J. F. Ru. Adaptive estimation and detection techniques with applications[D]. New Orleans, University of New Orleans, 2005.

[65] J. F. Ru, X. R. Li. Multiple-model detection of target maneuvers[C]//Proceedings of 2004 SPIE Conference on Signal and Data Processing of Small Targets, 2005, 5913: 1-9.

[66] J. F. Ru, V. P. Jilkov, X. R. Li, A. Bashi. Sequential detection of target maneuvers[C]//Proceedings of 2005 International Conference on Information Fusion, 2005, 345-351.

[67] Y. Baram, N. R. Sandell. An information theoretic approach to dynamical systems modeling and identification[J]. IEEE Transactions on Automatic Control, 1978, 23(1): 61-66.

[68] Y. Liu, X. R. Li. Performance analysis of Wald’s SPRT with independent but non-stationary log-likelihood ratios[C]//Proceedings 14th International Conference nformation Fusion, 2011, 145-152.

[69] Y. Bar-Shalom, X. R. Li, T. Kirubarajan. Estimation with applications to tracking and navigation: theory, algorithms, and software[M]. Boston, MA: Wiley, 2001.

[70] G. H. Jajamovich, A. Tajer, X. Wang. Minimax-optimal hypothesis testing with estimation-dependent costs[J]. IEEE Transactions on Automatic Control, 2012, 60(12): 6151-6165.

[71] A. Runnalls. Kullback-Leibler approach to Gaussian mixture reduction[J]. IEEE Transactions on Aerospace and Electronic Systems, 2007, 43(3): 989-999.

[72] X. R. Li, V. P. Jilkov, J. F. Ru. Multiple-model estimation with variable structure- part VI: expected-mode augmentation[J]. IEEE Transactions on Aerospace and Electronic Systems, 2005, 41(3): 853-867.

[73] J. Lan, X. R. Li, C. Mu. Best model augmentation for variable-structure multiple-model estimation[J]. IEEE Transactions on Aerospace and Electronic Systems, 2011, 47(3): 2008-2025.

[74] S. Ji, Z. Zhang, S. Ying, L. Wang, X. Zhao, Y. Gao. Kullback-Leibler divergence metric learning[J]. IEEE Transactions on Cybernetics, 2022, 52(4): 2047-2058.