Isochronal-band-constrained multipath structure matching for active sonar localization in convergence zones

In convergence-zone (CZ) propagation, sound rays undergo continuous refraction due to the vertical sound speed gradient, forming complex multipath arrival structures. Many conventional active-sonar localization methods approximately assume straight-line propagation and employ R = ct/2 for ranging. This simplification neglects ray bending and multipath superposition in CZ environments, thereby introducing systematic range-initialization bias that degrades localization accuracy. To address this issue, this paper proposes a multipath structure matching localization method based on two-way travel-time (TWTT) isochronal-band constraints. Using ray theory, a two-way propagation signal model for the ocean acoustic waveguide is derived. Matched filtering is employed to extract acoustic travel times, and an isochronal band centered on the first-arrival TWTT is constructed to replace the R = ct/2 linear formula for horizontal range correction. Subsequently, multipath arrival structure matching is used for depth estimation. To handle sound speed profile (SSP) mismatch, a feature-reliability adaptive weighting mechanism is introduced to determine the weighting coefficients in the matching cost function. Numerical simulations show that under a canonical Munk SSP, the proposed method reduces the depth error from 158 m to 36 m and keeps the range estimate at the grid-resolution level. Under SSP mismatch conditions with σc = 0 ∼ 3 m/s, TWTT-MSML reduces the mean depth error from 156–182 m to 36–81 m under the Munk SSP, and from 146–150 m to 25–76 m under the SCS WOA18 SSP. Weight ablation experiments indicate that the learned amplitude weight wAis close to the empirical optimum obtained from the ablation test. Elevation-angle ablation experiments show that the adaptively learned elevation-angle weight is substantially smaller than the relative-arrival-delay weight, indicating that elevation-angle features contribute only marginally to localization performance.