Bidirectional joint iteration adaptive estimation of underwater acoustic channel

Underwater acoustic (UWA) channel estimation is crucial for enhancing UWA communication. Conventional adaptive channel estimation techniques, relying on adaptive filtering algorithms, often falter under conditions of significant time variation, low signal-to-noise ratios (SNR), or extensive multipath delay spread. Drawing inspiration from bidirectional channel equalization, which mitigates error propagation via directional diversity, we propose the innovative adaptive filtering algorithms that simultaneously leverages bidirectional diversity through joint iteration. Specifically, the proposed algorithm exploits the time-reversal (TR) correlation relationship between the forward and reverse channel. By integrating coefficients from both forward and reverse adaptive filters and feeding them back as iterative initial values, we create a bidirectional joint iteration adaptive filtering framework compatible with conventional algorithms such as least mean square (LMS) and recursive least square (RLS). Meanwhile, we provide the theoretical derivation of optimal weight factors for both directions, maximizing diversity gains. Additionally, we present theoretical models detailing the transient behavior of the proposed algorithms. Numerical simulation and field experiment results demonstrate the superior performance of our bidirectional joint iteration adaptive filtering algorithms, particularly in challenging environments characterized by time-variation, impulsive noise, and low SNR. •Proposed Bi-Ji-LMS and Bi-Ji-RLS algorithms integrate a bidirectional joint iteration mechanism for improved low-SNR performance.•Derived optimal weight factors for bidirectional diversity gains and developed models analyzing Bi-Ji-LMS and Bi-Ji-RLS behavior.•Numerical simulations validate the theoretical analysis and evaluate the performance of our proposed algorithms.•Field experiments under high-noise conditions demonstrate the superiority and practical utility of the proposed algorithms.