Analysis of a Regular Tetrahedral Array for Underwater DOA Estimation

This paper presents a robust 3-D Direction of Arrival estimation framework using a compact regular tetrahedral array of four hydrophones, designed for shallow-water environments with adverse Signal-to-Noise Ratios (SNRs) down to −14 dB. The method combines GCC-PHAT for Time Difference of Arrival (TDOA) estimation with a closed-form Least Squares (LS) solver, enhanced by a recursive outlier-rejection mechanism to mitigate inconsistent TDOA pairs. Temporal coherence is exploited through multi-segment aggregation using mean and median operators. Simulations using synthetic signals generated by combining real recorded motorboat signatures with real underwater ambient noise demonstrate that discarding the two most inconsistent TDOA pairs reduces angular Root Mean Square Error (RMSE) by over 10° at −14 dB. Furthermore, median aggregation significantly outperforms mean aggregation; at a 5.12 s integration time and −14 dB SNR, the RMSE/Std improves from 13°/5° (LS-Mean, six pairs) to 2°/1° (LS-Median, four pairs). The results indicate that a 5.12 s integration window optimizes the trade-off between accuracy and update rate, proving suitable for real-time maritime situational awareness.