Evaluating underwater acoustics sensor network based on sparse LMS algorithm driven physical layer

Compared with the radio channels, underwater acoustic (UWA) channels pose challenging difficulties, such as the long time propagation, limited bandwidth, random multipath, and the doppler effect. By exploring the inherent sparsity caused by multipath structure, the reliability of underwater communication can be well improved through channel estimation and the equalization method in the physical layer. In this paper, a type of low-complexity channel estimation algorithms based on the least means square (LMS) adaptive iteration with different norm constraints are reviewed, like the l0-norm (l0-LMS), l1-norm (l1-LMS), and non-uniform norm (NNCLMS), from the perspective of the underwater network. The peer-to-peer long time-scale (in hours) channels variation observed in the physical shallow water channel is embedded into the Network Simulator 3 (NS3). The comprehensive performance of the UWA network was evaluated in terms of throughput, and end-to-end time delay. Lastly, an LSTM based Kalman filter (LSTM-KF) has been applied to predict channel response based on experimental NNCLMS estimation, which offers the potential to artificial extend the time scale of performance evaluation.