Channel knowledge map (CKM) has emerged as a crucial technology for next-generation communication, enabling the construction of high-fidelity mappings between spatial environments and channel parameters via electromagnetic information analysis. Traditional CKM construction methods like ray tracing are computationally intensive. Recent studies utilizing neural networks (NNs) have achieved efficient CKM generation with reduced computational complexity and real-time processing capabilities. Nevertheless, existing research predominantly focuses on single-antenna systems, failing to address the beamforming requirements inherent to MIMO configurations. Given that appropriate precoding vector selection in MIMO systems can substantially enhance user communication rates, this paper presents a TransUNet-based framework for constructing CKM, which effectively incorporates discrete Fourier transform (DFT) precoding vectors. The proposed architecture combines a UNet backbone for multiscale feature extraction with a Transformer module to capture global dependencies among encoded linear vectors. Experimental results demonstrate that the proposed method outperforms state-of-the-art (SOTA) deep learning (DL) approaches, yielding a 17% improvement in RMSE compared to RadioWNet. The code is publicly accessible at https://github.com/github-whh/TransUNet.

Underwater acoustic communication (UWA) faces significant coverage challenges due to the depth-varying sound speed gradients and the presence of sound shadow zones. Acoustic reconfigurable intelligent surface (RIS) is promising as an enabler to enhance acoustic signal quality and reliability. In this paper, we propose a novel full-phase-range acoustic RIS with effective acoustic beamforming scheme. Electrical unit parameters are carefully designed with Tonpilz hardware and equivalent circuit architecture. The reflective magnitude-phase coupling is analytically modelled by dual-quadratic expression. Furthermore, we propose one Majorization-Minimization (MM)-based acoustic RIS beamforming scheme, where alternative maximization approach is coordinated with fractional programming and MM methods to achieve the convex relaxation and near-optimal solutions.

Beyond 5G mobile networks are required to support the ultra-reliable data transmission with massive connectivity under high-mobility scenarios, where the delay-Doppler domain multiple access (DDMA) has been regarded as one of the potential candidates to avoid the severe performance degradation brought by double-dispersive channels. However, existing DDMA transceiver designs heavily rely on the shared codebooks or large guard space among user equipments (UEs), which restricts the flexibility, complexity, and spectral efficiency significantly. In this paper, we first propose the Zadoff-Chu training sequences-based frame structure and an element-wise iterative successive interference cancellation (SIC)-maximal ratio combining (MRC) detector for single-user transmission, which serves as the basis of flexible resource allocation in the delay-Doppler (DD) domain. The discussion is then extended to the MU downlink scenario, where rate splitting is adopted to effectively balance the noise and interference at the UE side to improve the bit error rate performance. For MU uplink cases, a non-orthogonal pilots-based iterative SIC-least square channel estimator is developed to promote the spectral efficiency while the iterative SIC-MRC detector is also provided. Simulation results demonstrate the excellent performance of the proposed DDMA scheme under typical DDMA patterns without guard space between UEs.

Communication objectives with high mobility bring severe Doppler shifts, causing the inter-carrier interference of orthogonal frequency division multiplexing (OFDM) system, which raises the requirements of novel modem schemes. However, existing modem schemes for high-mobility communications face challenges in adapting to diverse channel environments, involving complex channel estimation and etc. Fortunately, the potential of deep learning (DL) has been exploited in various communication applications. In order to design the consistent and robust modem scheme for different channel environments, we propose the DL-based architecture termed the closer twins (CTs) model, which borrows the idea from the Siamese structure in contrastive learning. In specific, two identical network backbones like twins can simultaneously process different channel inputs and make outputs consistent. We design a convlutional neural network called modem network (ModNet) as the backbone for the design of consistent and robust modem scheme. Moreover, to make traditional channel estimation and interpolation methods applicable to the designed modem scheme, a training-aided strategy called random-pilot (R-P) is proposed. In R-P strategy, we simulate the process of conventional channel estimation to modify the objective function of the modem scheme design. Furthermore, the performance of traditional channel estimation can be further improved by DL-based methods. We utilize the CTs model and design the backbone called estimation matrix network (EMNet) to optimize a linear channel estimation method, who outperforms the traditional methods with a similar complexity. Simulation results demonstrate that the proposed modem scheme outperforms OFDM, especially with high Doppler spread. The channel estimation strategy, supported by the R-P strategy and EMNet, achieves lower normalized mean square error compared with traditional methods, contributing to more reliable transmission.