Answers are grounded in this paper and cite their sources.
Near-field underwater perception requires spatially resolved measurements that remain informative in the presence of self-motion interference. In this paper we present a Dual-Modality Adaptive Artificial Lateral Line (DM-AALL) framework combining co-located SN-CN sensing, a bio-inspired layout and shared learning across hydrodynamic tasks. Superficial neuromast (SN) and canal neuromast (CN) channels provide complementary streamwise-velocity and pressure-difference observations at the same 16 positions. A spatially coherent adaptive periodic noise cancellation (SC-APNC) module exploits the spatial coherence of platform-induced interference, it removes only the portion of each channel estimate that is coherent with that reference, and this prevents the over-cancellation of target-band leakage that occurs when the fitted sinusoid is subtracted in full. Additionally, a physics-guided dual-modality fusion network (PGDM-Net) combines temporal features with bidirectional cross-attention and uses task-masked supervision to handle localization, ambient-flow estimation and wake inference within one model. Under an end-to-end simulation, the complete system achieves a mean radial localization error of 66.33 mm, flow-speed MAE of 0.172 m/s, and wake-frequency MAE of 0.721 Hz. A dedicated layout reassessment found that the longitudinal bio-inspired coordinates were statistically tied with an equally spaced array, whereas widening the lateral aperture from 12 to 48 mm lowered the mean localization error by about 24% before the gain plateaued.
No takes yet. Share an insight, caveat, or question.
You et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: