Data-Driven Modulation Optimization for Underwater Acoustic Communications With Joint Channel Estimation and LMMSE Equalization
Abstract
Ultra-reliable underwater acoustic (UWA) communications constitute a key enabling technology for future space–air–ground–underwater integrated networks. However, the reliability of current UWA transmissions remains limited due to the significant delay and Doppler scaling spread. Since interference-free transmission cannot be achieved under double spreads, we aim to enhance reliability by fully exploiting the modulation capability to reconstruct the interference with respect to the widely-used linear minimum mean square error (LMMSE) equalizer. The corresponding optimization problem is first reformulated as the design of a consistent modulation scheme that maximizes the fairness of the estimation mean square error (MSE) across all data symbols, leveraging the invariance of the total MSE induced by orthogonal modulation. The Siamese architecture is then employed to obtain a consistent modulation scheme across diverse channel conditions, which avoids the overhead of online feedback, transceiver cooperation, and network deployment. Furthermore, to enable practical transmission under the complicated channel structure induced by the proposed modulation framework, the superimposed training sequence is introduced. A joint orthogonal matching pursuit (OMP)-based channel estimation and LMMSE equalization scheme is then proposed to iteratively refine the estimated data symbols. Numerical results demonstrate that the proposed modulation scheme significantly outperforms alternatives under both perfect channel state information and practical transmission scenarios with the aid of the joint channel estimation and LMMSE equalization scheme.