Federated Learning (FL) has been widely used to train shared machine learning models while addressing the privacy concerns. When deployed in wireless networks, bandwidth resources limitation is a key issue, thereby necessitating device scheduling and bandwidth allocation. It is challenging to carry out device scheduling due to the large combinatorial search space. Besides, the heterogeneous computing capabilities and uncertain channel states of wireless devices complicate the design of a bandwidth allocation method. In this paper, we propose a joint device scheduling and bandwidth allocation framework for implementing FL in wireless networks. Specifically, deep reinforcement learning (DRL) is employed to conduct device scheduling. To this end, the state space, action space, and reward function of DRL are carefully defined for a typical FL system. Long short-term memory (LSTM) is adopted as the DRL agent to analyze the sequential input data. Given the scheduled devices of each global iteration, the proposed bandwidth allocation method aims to minimize the weighted sum of the time delay and energy consumption. Numerical experiments on both independent and identically distributed (IID) and non-IID datasets demonstrate that the proposed framework enables FL to reach the desired accuracy with low time delay and energy consumption.

Federated Learning (FL) is a promising machine learning approach for Internet of Things (IoT), but it has to address network congestion problems when the population of IoT devices grows. Hierarchical FL (HFL) alleviates this issue by distributing model aggregation to multiple edge servers. Nevertheless, the challenge of communication overhead remains, especially in scenarios where all IoT devices simultaneously join the training process. For scalability, practical HFL schemes select a subset of IoT devices to participate in the training, hence the notion of device scheduling. In this setting, only selected IoT devices are scheduled to participate in the global training, with each of them being assigned to one edge server. Existing HFL assignment methods are primarily based on search mechanisms, which suffer from high latency in finding the optimal assignment. This paper proposes an improved K-Center algorithm for device scheduling and introduces a deep reinforcement learning-based approach for assigning IoT devices to edge servers. Experiments show that scheduling 50% of IoT devices is generally adequate for achieving convergence in HFL with much lower time delay and energy consumption. In cases where reduction in energy consumption (such as in Green AI) and reduction of messages (to avoid burst traffic) are key objectives, scheduling 30% IoT devices allows a substantial reduction in energy and messages with similar model accuracy.

The registration of optical and synthetic aperture radar (SAR) images is crucial for multisource remote sensing image analysis and application. Due to different imaging mechanisms, the repeatable key points between optical and SAR images are scarce. Recent works have mainly focused on improving feature detection and description, but the matching strategies still rely on neighboring search, which only considers the visual appearance of key points while neglecting the spatial relationships. To resolve the issue, this letter proposes a two-stage registration algorithm based on combined features and graph neural networks (GNNs). First, to obtain more repeatable points, a convolutional neural network (CNN) is designed by combining detector-free and detector-based strategies, of which the former generates feature points with a grid-like distribution and the latter detects key points with a distinctive appearance. Second, GNNs are utilized for feature matching. The positional information of feature points is embedded into descriptors, and the information from other feature points is aggregated through an attention-based context aggregation mechanism to enrich feature descriptions. Third, a two-stage registration framework is adopted to raise the registration accuracy. Finally, the experimental results show that the proposed method performs excellently for the registration of optical and SAR images, maintaining a high matching success rate (SR) and accuracy under various conditions, including large-scale rotations, scale changes, and even perspective transformations.

The demand for high-resolution and wide-swath (HRWS) for squint azimuth multichannel synthetic aperture radar (MSAR) platform is increasingly urgent. However, the existence of interference will seriously contaminate the squint MSAR imagery; in particular, the deceptive forwarding interference (DFI) produced by the digital radio frequency memory (DRFM) technology makes the synthetic aperture radar (SAR) imagery confusing. For this point, the research on anti-DFI of squint MSAR is discussed in detail in this article. The presence of the Doppler ambiguity of the signal increases the complexity and difficulty of the DFI suppression. To solve this problem, the difference in the space–time spectrum between the valid signal and the DFI is first analyzed, and the steering vectors of each component in the echo are mined as prior information. Based on the prior information, a two-step processing is performed: the first step is to suppress the main-lobe DFI by using subspace projection and the second step is to suppress sidelobe DFI and complete signal spectrum reconstruction with the multiple Doppler direction linearly constrained minimum variance (MDD-LCMV) beamformer. Finally, the two experiment results show the excellent performance of the proposed method in squint MSAR for suppressing DFI.

Accurate range model with acceleration, the coupling phase terms, and spatial-variant (SV) Doppler parameters are the main issues to be solved in high-speed-high-squint SAR (HSHS-SAR) with a curved trajectory. For these issues, an extended Omega-K (EOK) algorithm is developed in this paper. The proposed EOK algorithm mainly includes the following four aspects. Firstly, a modified range model (AMRM) considering three-dimension acceleration for a curved trajectory is established. Then, the coupling between the range and azimuth direction is removed by the modified Stolt mapping (MSM). Subsequently, an improved high-order spatial-variant (SV) phase correction approach is derived to eliminate the azimuth dependence of Doppler parameters. Finally, in order to avoid zeros-padding operation, the proposed method focuses on the sub-aperture data in the range time and azimuth frequency domain through data aligning processing. The experimental results of both simulation and real data verify the effectiveness of the proposed method.

Machine learning (ML) is a widely accepted means for supporting customized services for mobile devices and applications. Federated Learning (FL), which is a promising approach to implement machine learning while addressing data privacy concerns, typically involves a large number of wireless mobile devices to collect model training data. Under such circumstances, FL is expected to meet stringent training latency requirements in the face of limited resources such as demand for wireless bandwidth, power consumption, and computation constraints of participating devices. Due to practical considerations, FL selects a portion of devices to participate in the model training process at each iteration. Therefore, the tasks of efficient resource management and device selection will have a significant impact on the practical uses of FL. In this paper, we propose a spectrum allocation optimization mechanism for enhancing FL over a wireless mobile network. Specifically, the proposed spectrum allocation optimization mechanism minimizes the time delay of FL while considering the energy consumption of individual participating devices; thus ensuring that all the participating devices have sufficient resources to train their local models. In this connection, to ensure fast convergence of FL, a robust device selection is also proposed to help FL reach convergence swiftly, especially when the local datasets of the devices are not independent and identically distributed (non-iid). Experimental results show that (1) the proposed spectrum allocation optimization method optimizes time delay while satisfying the individual energy constraints; (2) the proposed device selection method enables FL to achieve the fastest convergence on non-iid datasets.

Highly squinted synthetic aperture radar (HS-SAR) data focusing is a challenging task due to the heavy coupling between the range and azimuth, which would lead to the failure of the traditional imaging algorithms. In order to overcome these issues, a modified nonlinear chirp scaling (CS) algorithm for HS-SAR is proposed in this manuscript. First, traditional linear range walk correction (LRWC), range compressing (RC), secondary range compressing (SRC), range curvature correction (RCC) are employed as the range processing. Subsequently, a modulation phase factor (MPF) is introduced to weaken the influence of spatial variant (SV). After that, a high order SV correction phase (SVCP) approach is derived to eliminate the azimuth dependence of Doppler parameters. In addition, the sub-aperture data is focused on the range time and azimuth frequency domain by SPECAN processing to avoid numerous zeros-padding operations. Real data processing is adopted to prove the efficiency and validity of the proposed method.

Accurate 2-D frequency spectrum (2-D FS) with acceleration, two-way range coupling terms, and spatial-variant Doppler parameters are the main problems to be solved in translational-variant bistatic forward-looking synthetic aperture radar (SAR) (TV BFSAR) with curved trajectory. For these issues, a modified omega-K imaging algorithm is derived in this article. The maximum usage of 2-D FS based on the method of series reversion (MSR) is achieved by linear range cell migration correction, and 2-D FS is linearized in bistatic range by using high-order polynomial fitting. Then, a method of azimuth resampling is introduced to implement compensation of spatial-variant Doppler parameters. Different from other bistatic omega-K methods, our newly proposed method focuses on the small-aperture data in the azimuth frequency domain to avoid azimuth aliasing without padding zeros and uses the frequency focusing position to study the model of spatial-variant phase. Simulation results and real data verify the effectiveness of the proposed method.

In high-maneuverability bistatic forward-looking synthetic aperture radar (HMBF-SAR) imaging, the acceleration leads to an increased residual range curve and a deepened two-dimensional spatial variance of Doppler parameters, which cannot be processed by the traditional algorithms. To address these problems, this paper establishes a more accurate digital representation for HMBF-SAR model and investigates an extended azimuth nonlinear Chirp Scaling (EANLCS) imaging method. In the flowchart of this paper, we first propose a more precise slant range model with improved expansion coefficients, and defines the range and azimuth direction of HMBF-SAR imaging. Then, a novel fast reference point (i.e., azimuth and range reference point) selection method is proposed to analyze two-dimensional spatial variance of signal characteristics, which is used to construct a high order model of residual range cell migration and Doppler parameters. Based on above analysis, we put forward an advanced imaging algorithm of combining the Second-order keystone and extended azimuth nonlinear chirp scaling (EANLCS) to compensate the increased residual range curve and two-dimensional spatial variance of Doppler parameters. Finally, the effectiveness of the proposed HBMF-SAR method is verified by several numerical simulations and comparative studies based on both the simulated and raw data.

High computational complexity and phase errors (PEs) are the main limitations of time-domain (TD) synthetic aperture radar (SAR) imaging algorithms. Accelerated fast backprojection (BP) (AFBP) algorithm avoids interpolation through wavenumber spectrum connection and is an efficient fast TD imaging algorithm. In order to deal with the image defocusing problem caused by PEs effectively and ensure rapid imaging, a TD autofocus method is proposed in this article, which is based on generalized sharpness metrics and the AFBP imaging model. The autofocus method is divided into two stages. First, for each subaperture (SA), the PE estimation model is established in unified polar coordinate (UPC), where the strong-scattering range-cell pixels are chosen to reduce memory burden and avoid repetitive imaging. The PE estimation is converted into a nonconvex optimization problem. Then, the genetic algorithm (GA) and the maximizing-maximum-pixel-value (MMPV) method are used to estimate the PEs. Second, SA images’ matching and constant PE’s compensation are performed to eliminate the residual PEs. The full-aperture well-focused image is obtained by the coherent accumulation of SA images. The effectiveness of the proposed method is proven by the results of simulation and real SAR data processing.

Spatial-variant phase errors (PEs) are important factors which defocus the synthetic aperture radar (SAR) image. In time-domain SAR imaging, the exact calculation of instantaneous range is carried out to realize imaging. Accurate trajectory is the key to compensate spatial-variant PEs and ensure image focus. Thus, an improved time-domain autofocus method based on three-dimensional motion errors (3-D MEs) estimation is proposed in this article. First, an improved maximizing-maximum-pixel-value method is used to estimate nonspatial-variant PEs. Meanwhile, a theoretical explanation combined with$N$-dimensional Euclidean space is described. Then, residual PEs and wrapped PEs are discussed successively. A part-overlapped partitioning scheme for sub-block images (SBIs) and a wrapped-PE model are proposed for 3-D MEs estimation. Then, the estimation problem is turned into a mixed integer programming problem, which can be solved by the combination of genetic algorithm (GA) and Tikhonov regularization. Finally, the well-focused image is obtained through updated trajectory. The effectiveness of the proposed method is proven by results of simulation and real SAR data processing.

Phase gradient autofocus (PGA) is widely used in estimating residue phase error due to its efficient and robust. However, its precision severely relies on sample quality. In this paper, an expediting phase gradient autofocus algorithm is proposed to solve the above problem. First, we extract valid echo data area from the received contaminated data. Second, the optimization of the azimuth window size is presented. It gets rid of the limitation that the traditional PGA depends on the experience value. Third, the phase error can be fast calculated without numerous IFFT and zero padding which decrease computational complexity. Furthermore, the computational cost is derived in detail. Finally, Numerical simulation and raw SAR data demonstrate that the new method can achieve better performance than conventional PGA.