Robust operational GSTDOA localization of underwater black-box beacons: low-SNR arrival-time extraction and multi-reception fusion beyond the four-node estimator

Sun, Qin, Hao, Zhang, and Zhao (IEEE TGRS 2021) proposed the generalized second-order time difference of arrival (GSTDOA) method for localizing a sunken aircraft black box whose acoustic beacon suffers crystal-oscillator period drift. GSTDOA eliminates the unknown emission epoch and period by taking second-order differences of arrival times recorded at arbitrarily selected reception points along a ship trajectory, so that no clock synchronization or knowledge of the emission period is required. The published realization, however, is a minimal four-reception estimator: with four receptions the number of equations equals the number of unknown horizontal coordinates, the measurement-covariance weighting drops out of the maximum-likelihood objective, and reception-node selection is handled by a coordinate-descent (RELAX) search minimizing a horizontal-dilution-of-precision (HDOP) metric. The authors acknowledge three open directions: robust arrival-time extraction at low signal-to-noise ratio before GSTDOA processing; trajectory or path planning so that the search platform's motion produces precision-favorable reception geometry; and a global-optimization or multi-sample fusion scheme that exploits all available receptions rather than only four. The open problem is the statistically proper fusion of all receptions, fed by a low-SNR-robust arrival-time front end.

Background

After an aircraft crashes at sea, its flight recorders ("black boxes") are found acoustically: each carries an underwater locator beacon that emits an acoustic pulse roughly once per second, and a surface vessel or autonomous underwater vehicle (AUV, an unmanned robotic submarine) detects these pulses and estimates the beacon position on the seabed. Long-range underwater acoustic localization (UWAL) is dominated by time-of-arrival (TOA) and time-difference-of-arrival (TDOA) methods — which convert measured arrival times, or differences of arrival times between receptions at known positions, into range or range-difference equations — and by direction-of-arrival (DOA) methods, which estimate bearing from a hydrophone array and degrade with increasing target distance. Classical TOA/TDOA requires knowing when, or at what period, the beacon emits. In the cold, high-pressure deep-sea environment the beacon's crystal oscillator drifts, changing the pulse period by an unknown amount (the source paper's lake trial measured a 98 microsecond deviation from the nominal 1.0735 s period); period-assuming TDOA then produces large localization errors.

To remove the unknown emission epoch and period, Sun, Zhang, Zheng, Fu, and Zhao (IEEE JOE 2020) introduced the second-order time difference of arrival (STDOA): a difference of two TDOAs computed from arrival times $t_n$ recorded at reception indices $n$, which cancels the emission epoch and a constant period. Their STDOA, however, requires symmetric index spacing ($n_2-n_1=n_3-n_2$), which forces the corresponding spatial reception nodes to be symmetric in time — a rigid constraint that maps poorly onto realistic, irregularly maneuvering receiver platforms. The same group also developed an improved TDOA variant for the black-box problem (IEEE GRSL 2021).

The source paper (Sun, Qin, Hao, Zhang, Zhao, IEEE TGRS 2021) generalized STDOA to GSTDOA: the scalar $g(n_1,n_2,n_3)=\frac{t_{n_3}-t_{n_2}}{n_3-n_2}-\frac{t_{n_2}-t_{n_1}}{n_2-n_1}$ is defined for arbitrary reception indices $n_1<n_2<n_3$, cancels the unknown epoch and arbitrary per-period offset, and equals the corresponding second-order difference of propagation times $d_n=\|\mathbf{x}_t-\mathbf{x}_n\|_2/c$ (with $c$ the sound speed, $\mathbf{x}_t$ the target position, and $\mathbf{x}_n$ the receiver position at index $n$). For $N$ selected receptions the GSTDOA scalars of consecutive triples form an overdetermined-capable matrix observation model $\mathbf{G}_g\mathbf{d}-\mathbf{g}=\mathbf{e}_g$ with $N-2$ equations, a Gaussian noise covariance $\mathbf{Q}_g$, and a maximum-likelihood (MLE) cost $(\mathbf{G}_g\mathbf{d}-\mathbf{g})^T\mathbf{Q}_g^{-1}(\mathbf{G}_g\mathbf{d}-\mathbf{g})$. Localization precision is characterized by the horizontal dilution of precision (HDOP), the $2\times 2$ covariance of the horizontal position error obtained by first-order error propagation through the arrival-time, ship-position, sound-speed, and depth uncertainties. A coordinate-descent algorithm (RELAX) selects four reception indices minimizing HDOP, initialized from an STDOA fix. In simulation, tank, and lake experiments GSTDOA achieved lower root-mean-square error (RMSE) than STDOA and TDOA under period drift (e.g. 6.73 m vs 10.06 m and 47.17 m in simulation) at per-period runtimes of tens of milliseconds.

Crucially, the realized estimator used only four receptions per position fix: with $N=4$ the $N-2=2$ equations exactly match the two unknown horizontal coordinates (the beacon depth is taken from seabed bathymetry), so the covariance weighting drops out and the MLE reduces to solving the unweighted nonlinear system. The paper's own framework — the covariance matrix $\mathbf{Q}_g$ and the weighted objective for general $N$ — was therefore never instantiated beyond this minimal case, and the paper identifies three unresolved practical and methodological directions for future work.

Problem Statement

The source paper leaves the GSTDOA framework incomplete in three respects, which form one open problem: turning GSTDOA from a minimal four-reception estimator into an operational, robust localization system for drifting-period underwater beacons. The common objective is a GSTDOA-based search capability whose end-to-end localization accuracy and robustness are quantified, with the estimator fed by the front end's reliable arrival times and fusing all available receptions. Specifically:

(1) Front end: develop arrival-time extraction that remains reliable at low signal-to-noise ratio (SNR) — including detection of the beacon pulse and selection of the direct path in multipath — so that the arrival times fed into GSTDOA second-difference processing are trustworthy in the low-SNR conditions of long-range deep-water search.

(2) Reception geometry: design the search platform's trajectory (ship or AUV path planning under operational constraints such as obstacle avoidance and search-time limits) so that the motion itself produces reception nodes that improve GSTDOA localization precision.

(3) Estimation: develop and integrate the global optimization algorithm or multi-sample fusion scheme that uses all available receptions — more than the four receptions per position fix of the published realization — including the statistically proper covariance weighting of the correlated GSTDOA observations, and quantify the resulting estimator robustness (accuracy, sensitivity to arrival-time, ship-position, sound-speed, and depth uncertainties) relative to the four-reception baseline.

The open problem comprises component (3), together with component (1). Component (2) is retained as context rather than as a required deliverable.

References

1. Sibo Sun, Shasha Qin, Yu Hao, Guangpu Zhang, Chunhui Zhao, "Underwater Acoustic Localization of the Black Box Based on Generalized Second-Order Time Difference of Arrival (GSTDOA)", IEEE Transactions on Geoscience and Remote Sensing 59(9):7245-7255, 2021. DOI 10.1109/TGRS.2020.3032982, https://doi.org/10.1109/TGRS.2020.3032982
2. Sibo Sun, Xinyu Zhang, Ce Zheng, Jin Fu, Chunhui Zhao, "Underwater Acoustical Localization of the Black Box Utilizing Single Autonomous Underwater Vehicle Based on the Second-Order Time Difference of Arrival", IEEE Journal of Oceanic Engineering 45(4):1268-1279, 2020. DOI 10.1109/JOE.2019.2950954, https://doi.org/10.1109/JOE.2019.2950954
3. Sibo Sun, Chen Zhao, Ce Zheng, Chunhui Zhao, Yilin Wang, "High-Precision Underwater Acoustical Localization of the Black Box Based on an Improved TDOA Algorithm", IEEE Geoscience and Remote Sensing Letters 18(8):1317-1321, 2021. DOI 10.1109/LGRS.2020.3002169, https://doi.org/10.1109/LGRS.2020.3002169
4. Sibo Sun, Tieyuan Liu, Yilin Wang, Guangpu Zhang, Kaixin Liu, Yong Wang, "High-Rate Underwater Acoustic Localization Based on the Decision Tree", IEEE Transactions on Geoscience and Remote Sensing 60:1-12, 2022. DOI 10.1109/TGRS.2021.3127919, https://doi.org/10.1109/TGRS.2021.3127919
5. Sibo Sun, Huigong Guo, Guangming Wan, Chao Dong, Ce Zheng, Yong Wang, "High-Precision Underwater Acoustic Localization of the Black Box Utilizing an Autonomous Underwater Vehicle Based on the Improved Artificial Potential Field", IEEE Transactions on Geoscience and Remote Sensing 61:1-10, 2023. DOI 10.1109/TGRS.2023.3249199, https://doi.org/10.1109/TGRS.2023.3249199
6. Xueli Sheng, Hang Dong, Xian Xiu, Yuan Gao, Kejing Sun, "Underwater target time difference of arrival estimation using low-rank decomposition based generalized cross correlation", The Journal of the Acoustical Society of America 160(1):658-670, 2026. DOI 10.1121/10.0044381, https://doi.org/10.1121/10.0044381
7. Bruno M. Ferreira, Paula A. Graça, José Carlos Alves, Nuno A. Cruz, "Single Receiver Underwater Localization of an Unsynchronized Periodic Acoustic Beacon Using Synthetic Baseline", IEEE Journal of Oceanic Engineering 48(4):1112-1126, 2023. DOI 10.1109/JOE.2023.3275611, https://doi.org/10.1109/JOE.2023.3275611
8. Yang Liu, Fucheng Guo, Le Yang, Wenli Jiang, "Source localization using a moving receiver and noisy TOA measurements", Signal Processing 119:185-189, 2016. DOI 10.1016/j.sigpro.2015.07.029, https://doi.org/10.1016/j.sigpro.2015.07.029
9. Francois Quitin, Zahra Madadi, Wee Peng Tay, "RF transmitter geolocation based on signal periodicity: Concept and implementation", 2015 IEEE International Conference on Communications (ICC), pp. 4593-4598, 2015. DOI 10.1109/ICC.2015.7249047, https://doi.org/10.1109/ICC.2015.7249047