A Hierarchical Adaptive Moment Matching Multiple Model Tracking Method for Hypersonic Glide Target Under Measurement Uncertainty

Hanxing Shao, Jibin Zheng, Yulong Bai, Hongwei Liu, Y. L. Ge, Boyang LiuSensors 2025 · DOI

TL;DR

This paper introduces a novel tracking method, HAMM-MEE, designed to accurately track hypersonic glide targets (HGTs) even with complex maneuvers and uncertain measurement noise. By combining a hierarchical adaptive model set with a robust filtering technique, it significantly improves tracking accuracy and stability compared to existing methods.

What Problem Does This Paper Address?

Hypersonic glide targets (HGTs) are notoriously difficult to track for several reasons:

  • High Speed and Complex Maneuvers: HGTs can execute a wide range of unpredictable maneuvers during their flight, making it hard for traditional tracking systems to keep up.
  • Measurement Uncertainty: Radar measurements can be noisy and their statistical properties can change over time, especially during atmospheric reentry.
  • Limitations of Existing Methods:
    • Single-model trackers fail to capture the diverse maneuvers.
  • Multiple-model (MM) methods often struggle with a trade-off between having enough models to cover all possibilities (leading to computational burden and model competition) and having too few (leading to poor performance when the target's true behavior isn't represented).
  • Many MM methods assume Gaussian noise, which is often not the case for HGTs, leading to degraded performance.

This paper aims to bridge these gaps by proposing a method that can adapt its models to the target's behavior and robustly handle non-Gaussian noise.

Methods

The proposed method, Hierarchical Adaptive Moment Matching Minimum Error Entropy Cubature Kalman Filter (HAMM-MEE), has two main components:

  1. Hierarchical Adaptive Moment Matching (HAMM) Multiple-Model Framework:
    • Comprehensive Model Set: First, a broad set of kinematic models is created to cover various HGT maneuver modes (e.g., no maneuver, random maneuver, periodic maneuver).
  2. Hierarchical Structure: This set is organized into two tiers:
  3. Coarse Model Set: This set is dynamically adapted. Underperforming models are terminated, and new candidates are activated based on Rényi divergence and posterior probability evolution. This allows the system to shed irrelevant models and adopt new ones as needed.
  4. Fine Model Set: This set is generated using a moment-matching method. It aims to capture the target's motion characteristics more precisely by matching the first two moments of the expected motion mode. This is done dynamically at each time step.
  5. Adaptive Strategy: This dual-layer approach allows for quick adaptation to maneuver changes while maintaining computational efficiency.
  6. Minimum Error Entropy Cubature Kalman Filter (MEECKF):
    • Robustness to Non-Gaussian Noise: To handle non-Gaussian measurement noise (like heavy-tailed or impulsive noise), the paper proposes using the Minimum Error Entropy (MEE) criterion. MEE is known for its robustness in such scenarios.
  7. Cubature Kalman Filter (CKF) Integration: The MEE criterion is integrated into a Cubature Kalman Filter (CKF) framework. CKF is chosen for its good balance of accuracy and computational cost compared to other extensions of the Kalman filter.
  8. Regression Model and SVD: To ensure stability, especially with potentially ill-conditioned covariance matrices due to noise, a regression model is constructed using statistical linearization. Singular Value Decomposition (SVD) is employed instead of Cholesky decomposition to enhance algorithmic stability.

Datasets

The paper uses simulated data based on the Lockheed Martin Common Aero Vehicle (CAV) model. Three specific trajectory scenarios are generated:

  1. Quasi-equilibrium glide trajectory.
  2. Jump glide trajectory.
  3. Penetration maneuver trajectory.

These scenarios are designed to represent typical HGT flight paths and are used to evaluate the performance of the proposed method under various noise conditions and sampling rates. No new benchmark datasets are introduced.

Key Insights and Ideas

  • Hierarchical Adaptation is Key: The paper demonstrates that a hierarchical approach to managing multiple models (coarse for broad adaptation, fine for precise matching) is more effective than a flat MM structure. This allows for both rapid response to maneuvers and efficient computation.
  • Rényi Divergence for Model Selection: Using Rényi divergence to adapt the coarse model set is a novel way to dynamically prune and select models, improving the system's ability to focus on relevant behaviors.
  • Moment Matching for Fine-Tuning: Generating a fine model set via moment matching allows for a more accurate representation of the target's current motion, complementing the broader adaptation of the coarse set.
  • MEE for Non-Gaussian Robustness: Integrating the Minimum Error Entropy criterion into a Cubature Kalman Filter is a powerful strategy for dealing with the challenging non-Gaussian noise often encountered in HGT tracking. This is a significant improvement over standard Kalman filters.
  • SVD for Stability: The use of Singular Value Decomposition in the filtering process enhances numerical stability, which is crucial when dealing with potentially ill-conditioned covariance matrices arising from noisy measurements.

Key Results

The proposed HAMM-MEE method consistently outperformed existing methods (IMM, BMA, and HAMM-CKF) in various simulation scenarios:

  • Improved Accuracy: HAMM-MEE achieved significantly lower Root Mean Square Errors (RMSE) and Average Root Mean Square Errors (ARMSE) across position, velocity, and acceleration estimates for all three trajectories.
  • Robustness to Noise: The method showed superior performance under Gaussian, heavy-tailed, and impulsive noise conditions, demonstrating its strong ability to handle non-Gaussian disturbances.
  • Adaptability to Sampling Rates: HAMM-MEE maintained low errors even when sampling rates varied, indicating its resilience to irregular data acquisition.
  • Computational Efficiency: While not the fastest, HAMM-MEE offered a good balance between performance and computational cost, being more efficient than IMM and comparable to BMA, making it suitable for real-time applications. HAMM-CKF was the fastest but less robust to non-Gaussian noise.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that while HAMM-MEE is robust, the performance degradation under severe non-Gaussian noise is still present, though less pronounced than in other methods.
  • The computational cost, while improved, can still be a factor for extremely high-dimensional problems or very rapid maneuver changes.
  • Future Work:
    • Investigate parallel computing architectures to further enhance computational speed.
  • Explore robust data association strategies for varying Signal-to-Noise Ratios (SNR).
  • Address scenarios involving missed detections.
  • Extend the framework to handle more complex scenarios, such as varying SNR and missed detections.

Summary updated

Abstract

Hypersonic glide targets (HGTs) pose significant challenges for radar tracking due to complex maneuver strategies and time-varying statistics of measurement noise. Conventional single-model tracking methods are generally insufficient to fully capture maneuver modes, while existing multiple-model methods face trade-offs between model set completeness and computational efficiency. In addition, existing tracking methods struggle to cope with the non-Gaussian noise during hypersonic flight. To overcome these limitations, a Hierarchical Adaptive Moment Matching (HAMM) multiple-model method is proposed in this paper. Firstly, a comprehensive model set is constructed to cover characteristic maneuver modes. Subsequently, a hierarchical multiple-model framework is developed where: (1) a coarse model set is dynamically adapted by multi-frame posterior probability evolution and Rényi divergence criteria; (2) a fine model set is generated based on the moment matching method. Furthermore, the minimum error entropy cubature Kalman filter (MEECKF) is proposed to suppress the non-Gaussian measurement noise with high stability. Monte Carlo simulations demonstrate that the proposed method achieves improved positioning accuracy and faster convergence.
BibTeX
@inproceedings{shao2025hierarchical,
  author = {Hanxing Shao and Jibin Zheng and Yulong Bai and Hongwei Liu and Y. L. Ge and Boyang Liu},
  title = {{A Hierarchical Adaptive Moment Matching Multiple Model Tracking Method for Hypersonic Glide Target Under Measurement Uncertainty}},
  booktitle = {Sensors},
  year = {2025},
  doi = {10.3390/s25216621},
}
Figure