This paper presents an end-to-end deep learning framework in a movable antenna (MA)-enabled multiuser communication system. In contrast to the conventional works assuming perfect channel state information (CSI) for MA placement or adopting a decoupled CSI acquisition and MA placement design paradigm, we address the practical CSI acquisition issue through the design of pilot signals and quantized CSI feedback, and further incorporate the joint optimization of channel estimation, MA placement, and precoding design. The proposed mechanism enables the system to learn an optimized transmission strategy from imperfect channel data, overcoming the limitations of conventional methods that conduct channel estimation and antenna position optimization separately. To balance the performance and overhead, we further extend the proposed framework to optimize the antenna placement based on the statistical CSI. Simulation results demonstrate that the proposed approach consistently outperforms traditional benchmarks in terms of achievable sum-rate of users, especially under limited feedback and sparse channel environments. Notably, it achieves a performance comparable to the widely-adopted gradient-based methods with perfect CSI, while maintaining significantly lower CSI feedback overhead. These results highlight the effectiveness and adaptability of learning-based MA system design for future wireless systems.

Multiband fusion (MF) technology can greatly enhance the range resolution of radar by generating ultrawideband echo using several subbands echoes (SBEs). A critical issue of MF is how to accurately estimate and compensate for incoherence between SBEs. In this letter, a novel high-accuracy incoherence compensation method is proposed. First, the phase-coherent coefficient (PCC) is introduced to quantify the phase correlation between SBEs, which is not exactly analyzed in the existing MF methods. Then, it is noticed that for PCC, the linear phase and the fixed phase are coupled and difficult to be estimated simultaneously. Accordingly, this letter proposes to estimate the linear phase and the fixed phase separately, where the linear phase is estimated and compensated by applying the information entropy minimization criterion, and the fixed phase is then compensated based on PCC. Finally, the validity and advantages of the proposed method are investigated by numerous simulation results.

The increasing popularity of Radar High-Resolution Range Profiles (HRRP) in radar automatic target recognition (RATR) is increasingly evident, primarily due to its simple imaging techniques and rapid processing capabilities. HRRP has emerged as a critical area of focus in contemporary research. Nonetheless, a significant obstacle in RATR is the impaired recognition performance caused by the absence of target-aspect from non-cooperative targets. To address this challenge, this paper presents an innovative approach, termed "Proxy-based Feature Ensemble Supervised Contrastive Learning." This method skillfully employs proxy-based contrastive learning to extract invariant feature representations. Additionally, the incorporation of the Geometrical Theory of Diffraction model for parametric feature representation, along with feature-level ensembling, substantially enhances model accuracy through feature ensembling. Simulated experiments reveal that in scenarios lacking target-aspect information, the proposed methodology achieves remarkable generalization ability in zero-shot tasks.

In the evolving realm of Radar Automatic Target Recognition utilizing High-Resolution Range Profiles, this study confronts a critical issue: the substantial impact of rapid attitude angle alterations on the accuracy of recognizing highly maneuverable targets. Regarding this issue, a novel self-supervised training method, named Contrastive-based Time-Frequency Domain Dynamic Fusion, is proposed, notably boosting generalization in zero-shot classification tasks. Initially, a novel data augmentation technique is proposed to combine with contrastive learning, thereby enhancing the robustness of time-domain features while effectively guiding the encoding of frequency-domain features. Furthermore, to streamline the dynamic integration of time-frequency domain features, feature alignment, and feature fusion subtasks based on contrastive learning are meticulously designed, fusing different features with a learning paradigm. Experiments evidence indicates that this method significantly surpasses existing models in scenarios with complex target maneuvers.

Positioning accuracy and efficiency are two important features for embedded positioning devices of Internet of Things (IoT) in smart cities. In modern urban canyon environments, Non Line-of-Sight (NLoS) satellite signals may degrade positioning accuracy. The mirror ray tracing algorithm can be used to reconstruct NLoS propagation path and enhance positioning accuracy. However, for the embedded positioning devices of IoT, real-time reconstruction of NLoS propagation path is difficult. In this article, an inherited sampling algorithm ensuring positioning accuracy and efficiency is proposed. Experimental results show that the proposed inherited sampling algorithm saves computation time by 69.8% compared to the double sampling algorithm. Furthermore, a mirror ray tracing accelerator is designed and built in FPGA, which greatly reduces the NLoS path reconstruction time. Finally, a real-time positioning system with a CPU and FPGA heterogeneous architecture is built. For a single trajectory point, test results show that the correction time of the proposed inherited sampling algorithm with an FPGA accelerator takes only 0.45 s, which is 80 times faster than that with CPU calculation alone.

In this article, we investigate a resource allocation problem for a multiunmanned aerial vehicle (UAV) assisted integrated sensing and communication (ISAC) system, where a group of dual-functional UAVs perform simultaneous radar sensing of a target and data communication with multiple ground users (GUs). In particular, the trajectory of UAVs, user association, and beamforming design are jointly considered to maximize the sum weighted bit rate of all GUs while ensuring the sensing beampattern gain of the target. To cope with the above mixed-integer nonconvex optimization problem, we propose an efficient strategy by decomposing the original problem into two subproblems under the alternating optimization framework. For the user association and beamforming design, we propose a novel algorithm to circumvent the coupling relationship among GUs and UAVs by leveraging matching theory and fractional programming theory. For the nonconvex UAV trajectory subproblem, we apply the sequential quadratic programming to obtain a suboptimal solution by solving a sequence of quadratic programming problems. The above two subproblems are iteratively solved and a stable solution is obtained upon convergence. Simulation results show that the proposed strategy outperforms various benchmark schemes that are based on the deferred acceptance algorithm, K-means algorithm, and a heuristic algorithm. It is demonstrated that the proposed strategy efficiently improve the sensing beampattern gain and communication rate.

Dual-frequency radar is a preferred solution for moving target localization because of its low complexity and cost. Although several studies for dual-frequency radar are addressed in the literature, all of those methods are designed for the synchronous mode. In this letter, a new localization method based on phase compensation is proposed for dual-frequency radar in the asynchronous mode. Compared with the synchronous mode, the proposed method is suitable for most commercial short-range radars with the single local oscillator architecture and can provide a cost-effective solution. Moreover, both the theoretical variance and Cramer-Rao lower bound are derived for the performance analysis in the asynchronous mode. Computer simulations were conducted to verify the validity of the proposed method.