To address the challenges of obtaining network state information, flexibly forwarding data, and improving the communication quality of service (QoS) in wireless network transmission environments in response to dynamic changes in network topology, this paper introduces an intelligent routing algorithm based on deep reinforcement learning (DRL) with network situational awareness under a software-defined wireless networking (SDWN) architecture. First, comprehensive network traffic information is collected under the SDWN architecture, and a graph convolutional network-gated recurrent unit (GCN-GRU) prediction mechanism is used to perceive future traffic trends. Second, a proximal policy optimization (PPO) DRL-based data forwarding mechanism is designed in the knowledge plane. The predicted network traffic matrix and topology information matrix are treated as the DRL environment, while next-hop adjacent nodes are treated as executable actions, and action selection policies are designed for different network conditions. To guide the learning and improvement of the DRL agent’s routing strategy, reward functions of different forms are designed by utilizing network link information and different penalty mechanisms. Additionally, importance sampling steps and gradient clipping methods are employed during gradient updating to enhance the convergence speed and stability of the designed intelligent routing method. Experimental results show that this solution outperforms traditional routing methods in network throughput, delay, packet loss rate, and wireless node distance. Compared to value-function-based Dueling Deep Q-Network (DQN) routing, the convergence of the proposed method is significantly faster and more stable. Simultaneously, hardware storage consumption is reduced, and real-time routing decisions can be made using the current network state information. The source code can be accessed at https://github.com/GuetYe/DRL-PPONSA.

Affected by the complex underwater environment and the limitations of low-resolution sonar image data and small sample sizes, traditional image recognition algorithms have difficulties achieving accurate sonar image recognition. The research builds on YOLOv7 and devises an innovative fast recognition model designed explicitly for sonar images, namely the Dual Attention Mechanism YOLOv7 model (DA-YOLOv7), to tackle such challenges. New modules such as the Omni-Directional Convolution Channel Prior Convolutional Attention Efficient Layer Aggregation Network (OA-ELAN), Spatial Pyramid Pooling Channel Shuffling and Pixel-level Convolution Bilat-eral-branch Transformer (SPPCSPCBiFormer), and Ghost-Shuffle Convolution Enhanced Layer Aggregation Network-High performance (G-ELAN-H) are central to its design, which reduce the computational burden and enhance the accuracy in detecting small targets and capturing local features and crucial information. The study adopts transfer learning to deal with the lack of sonar image samples. By pre-training the large-scale Underwater Acoustic Target Detection Dataset (UATD dataset), DA-YOLOV7 obtains initial weights, fine-tuned on the smaller Smaller Common Sonar Target Detection Dataset (SCTD dataset), thereby reducing the risk of overfitting which is commonly encountered in small datasets. The experimental results on the UATD, the Underwater Optical Target Detection Intelligent Algorithm Competition 2021 Dataset (URPC), and SCTD datasets show that DA-YOLOV7 exhibits outstanding performance, with mAP@0.5 scores reaching 89.4%, 89.9%, and 99.15%, respectively. In addition, the model maintains real-time speed while having superior accuracy and recall rates compared to existing mainstream target recognition models. These findings establish the superiority of DA-YOLOV7 in sonar image analysis tasks.

In this paper, we aim at the problem that MIMO radar’s target detection performance is greatly reduced in the complex multi-signal-dependent interferences environment. We propose a joint design method based on semidefinite relaxation (SDR), fractional programming and randomization technique (JD-SFR) and a joint design method based on coordinate descent (JD-CD) to solve the actual transmit waveform and receive filter bank directly to reduce the loss of strong interference to the output signal-to-interference-plus-noise ratio (SINR) of the radar system. Therefore, the maximization of output SINR is taken as the criterion of the optimization problem. The designed waveforms take into account the radar transmitter’s hardware requirements for constant envelope waveforms and impose similarity constraints on the waveforms. JD-SFR uses SDR, fractional programming and randomization technique to deal with the non-convex optimization problems encountered in the solution process. JD-CD transforms the optimization problem into a function of the phase of the waveform and then solves the transmit waveform based on CD. Compared with other methods, the proposed method has lower output SINR loss under strong power interference and forms deep nulls on the direction beampattern of multiple interference sources, which indicates that it has better anti-interference performance.

For user terminals, energy-detection based multiple-input-multiple-output (MIMO) has the advantages of low complexity and anti-Doppler capability. However, the energy detector makes it difficult to linearly separate the signals in spatial multiplexing. In this letter, the linearized model is introduced in $2\times 2$ M-ary frequency-shifted keying (MFSK) V-BLAST system with energy detection. To avoid rank deficiency of the model, we propose nonlinear virtual antennas (NVA) to increases both the dimension and receive diversity gain of MIMO system. Then the full-rank condition is discussed. In multiple-antenna systems, NVA is further generalized by spatial index modulation with dual antennas selection. By simulations, linear detectors are verified to be feasible for non-coherent spatial-multiplexing MIMO, and its diversity gain is obviously increased by the proposed NVA.