Underwater target classification has been an important topic driven by its general applications. Convolutional neural network (CNN) has been shown to exhibit excellent performance on classifications especially in the field of image processing. However, when applying CNN and related deep learning models to underwater target classifications, the problems, including small sample size of underwater target and low complexity requirement, impose a great challenge. In this paper, we have proposed the modified DCGAN model to augment data for targets with small sample size. The data generated from the proposed model help to improve classification performance under imbalanced category conditions. Furthermore, we have proposed the S-ResNet model to obtain good classification accuracy while significantly reducing complexity of the model, and achieve a good tradeoff between classification accuracy and model complexity. The effectiveness of proposed models is verified through measured data from sea trial and lake tests.

In this letter, we first present a time-domain model for pilot-assisted MMSE channel estimation in the orthogonal chirp division multiplexing (OCDM) systems. Then by proposing a pilot selection criterion, the complexity of MMSE channel estimation is reduced by the simplification of matrix inversion. Meanwhile, we prove that the criterion can lead to the optimal mean square error (MSE) performance. Furthermore, based on the criterion, by adopting the Zadoff-Chu sequences as pilots and presenting a split-and-decomposition algorithm, we propose a low-complexity channel estimator for OCDM. The complexity of the proposed estimator is O (Nlog\frac L2), while the estimator in existing literature for OCDM is O (NlogN), where L is the channel length and N is the number of subcarriers. Numerical results are provided to verify the performance of the proposed algorithm over frequency-selective Rayleigh fading channels. It turns out that the proposed estimator can significantly reduce the computational complexity without any MSE performance loss.

The channel reservation mechanism is devised to reduce end-to-end delay and improve throughput for underwater acoustic sensor networks (UASNs) with dynamic load. Unfortunately, in dynamic-high-load UASNs, frequent channel reservation will lead to an increase in end-to-end delay and a decrease in throughput due to long propagation delay and collisions. To reduce end-to-end delay and improve throughput, we proposed an adaptive MAC protocol for UASNs with dynamic-high load (DHL-MAC). DHL-MAC performs channel reservation before the data packets are actually generated and avoids collisions to reduce the impact of channel reservation on end-to-end delay and throughput. First, the adaptively allocated time slot for the next time frame is obtained based on the estimation factor and secondary transmission data packets, which could reduce idle waiting time. Then, by merging the channel reservation information for next time frame into the current data frame, channel reservation is performed before data packet generation to reduce waiting delay. Finally, a dynamic time slot schedule scheme is proposed to reduce time frame length and achieve collision avoidance. The performance analysis and simulation results show that our proposed DHL-MAC could efficiently reduce end-to-end delay and improve throughput in dynamic-high-load UASNs. Simulation results show that DHL-MAC can achieve a maximum 95.2% reduction in average end-to-end delay and a maximum 56.2% improvement in throughput upper bound compared with R-MAC.

Many marine scientific and industrial applications in underwater acoustic sensor networks (UASNs) are characterized by dynamic-high load and dense deployment. In such UASNs, the channel reservation mechanism is devised to deal with channel utilization degradation. Unfortunately, channel reservation would lead to an increase in end-to-end delay and a decrease in throughput due to long propagation delay in UASNs. To reduce end-to-end delay and improve throughput, we proposed an efficient distributed MAC protocol for UASNs (ED-MAC). ED-MAC reduces the impact of long propagation delay on channel reservation by removing the channel reservation confirmation process and improving channel reservation efficiency. First, the optimized time slot length for the next time frame is obtained based on data packet number estimate and secondary transmission data packet number, which could reduce idle waiting time. Then, the channel reservation confirmation process is removed by distributed channel reservation mechanism, and efficient channel reservation is achieved through channel reservation packet collision-free transmission and channel multiplexing. Meanwhile, channel reservation is performed before packet generation. Thus, data packet waiting delay is reduced. Finally, a distributed data packet time slot schedule scheme is proposed to achieve collision avoidance between different data packets as well as data packets and channel reservation packets. The performance analysis and simulation results show that our proposed ED-MAC could efficiently reduce end-to-end delay and improve throughput in dynamic-high-load and dense UASNs. Simulation results show that ED-MAC can achieve a maximum 98.7% reduction in average end-to-end delay and a maximum 61.1% improvement in throughput.

Long-term underwater acoustic sensor networks (UASNs) are widely deployed for marine scientific and industrial applications. The widespread ocean currents lead to variations in the communication range of location-fixed nodes equipped with non-omnidirectional transducers, which will reduce the successful packet relaying probability. In order to reduce the impact of ocean current on packet relaying, we propose an adaptive routing protocol for underwater acoustic sensor networks with ocean current (OCAR). By adaptively selecting candidate nodes based on communication range, OCAR improves the successful packet relaying probability in poor communication conditions caused by ocean current. Firstly, the impact of ocean current on communication range is demonstrated by constructing a communication range model and a sea trial. Secondly, successful packet relaying probability is improved by selecting the communicable nodes under ocean current as the candidate nodes. Next, the transmitter-based adaptive routing scheme reduces the number of nodes responding to the relay packet and adaptively reselects the destination node per hop. Therefore, energy consumption is reduced and the packet delivery ratio (PDR) is improved. Finally, routing recovery and candidate set dynamic update are achieved to improve successful packet relaying probability. Simulation results show that, compared with Flood, OCAR can achieve a maximum 98.29% reduction in energy consumption and 37.05% reduction in average end-to-end delay while maintaining close to 100% PDR. In addition, compared with VBF, OCAR can achieve close to 100% PDR when the PDR of VBF is 0, maximum energy consumption reduction of 84.41% and average end-to-end delay reduction of 96.34%.

The unique characteristic of bandwidth-distance relationship in underwater acoustic (UWA) channels makes standard uplink non-orthogonal multiple access (NOMA) scheme not be applicable in general underwater acoustic networks (UWANs). In this article, we propose an uplink partial overlapping non-orthogonal multiple access (PONOMA) scheme for UWANs, and investigate capacity region of uplink PONOMA in UWANs. By formulating the problem of the optimal energy distribution for uplink PONOMA based on UWA channel modeling, the closed-form solutions for optimal power spectral density (p.s.d.) and capacity region are derived. Furthermore, we modify the channel modeling by building in parameter uncertainty induced by the empirical nature of UWA channel modeling using box sets, and derived robust capacity region under parameter uncertainty. Numerical results are provided to demonstrate the performance of our proposed scheme.