11 September 2020 in IEEE Access
Recognizing the various and broad range of applications of unmanned aerial vehicles (UAVs) and unmanned aircraft systems (UAS) for personal, public and military applications, recent un-intentional malfunctions of uncontrollable UAVs or intentional attacks on them divert our attention and motivate us to devise a protection system, referred to as a counter UAV system (CUS). The CUS, also known as a counter-drone system, protects personal, commercial, public, and military facilities and areas from uncontrollable and belligerent UAVs by neutralizing or destroying them. This paper provides a comprehensive survey of the CUS to describe the key technologies of the CUS and provide sufficient information with wich to comprehend this system. The first part starts with an introduction of general UAVs and the concept of the CUS. In the second part, we provide an extensive survey of the CUS through a top-down approach: i) the platform of CUS including ground and sky platforms and related networks; ii) the architecture of the CUS consisting of sensing systems, command-and-control (C2) systems, and mitigation systems; and iii) the devices and functions with the sensors for detection-and-identification and localization-and-tracking actions and mitigators for neutralization. The last part is devoted to a survey of the CUS market with relevant challenges and future visions. From the CUS market survey, potential readers can identify the major players in a CUS industry and obtain information with which to develop the CUS industry. A broad understanding gained from the survey overall will assist with the design of a holistic CUS and inspire cross-domain research across physical layer designs in wireless communications, CUS network designs, control theory, mechanics, and computer science, to enhance counter UAV techniques further.
9 April 2019 in IEEE Communications Letters
In this letter, the altitude of an unmanned aerial vehicle base station (UAV-BS) is optimized to maximize an average worst case secrecy rate (AWSR) without eavesdropper location information. The air-to-ground channels are modeled based on the derived line-of-sight probability under a general urban model with Poisson point process and Rayleigh distribution for the locations and heights of the buildings, respectively. To solve the AWSR maximization problem, which is intractable, we reformulate it by the first-order Taylor series approximation and apply the Davies–Swann–Campey algorithm to effectively find the solution of the modified AWSR maximization problem. Numerical results verify that the altitude of UAV-BS affects the secrecy rate significantly and that the proposed UAV-BS can improve the secrecy rate compared with a static terrestrial BS.
7 July 2021 in IEEE Transactions on Vehicular Technology
In this correspondence, we study a tethered unmanned aerial vehicle (TUAV)-assisted mobile edge computing (MEC) system, in which tasks of several ground mobile users (MUs) are served by ground edge cloud (GEC) with superior computing capability and poor communication links and TUAV-mounted cloudlet with lower computing capability and better communication links. We aim to minimize weighted-sum system delay of MUs by jointly optimizing TUAV position, time slot allocation, and task splitting ratio. We first decompose the optimization problem into three subproblems, and then to solve non-convex subproblem related to TUAV position, we reformulate it by the first-order Taylor series approximation and apply the successive convex approximation (SCA) method. Then, we tackle the overall problem by applying the alternative optimization algorithm. Numerical results verify that our proposed cooperative sky-ground mobile edge computing system (CoSMoS) algorithm can achieve a better performance than other benchmark schemes.
10 March 2025 in IEEE Transactions on Mobile Computing
Federated learning (FL) enables distributed training while preserving data privacy, but stragglers—slow or incapable clients can significantly slow down the total training time and degrade performance. To mitigate the impact of stragglers, system heterogeneity, including heterogeneous computing and network bandwidth, has been addressed. While previous studies have addressed system heterogeneity by splitting models into submodels, they offer limited flexibility in model architecture design, without considering potential inconsistencies arising from training multiple submodel architectures. We propose nested federated learning (NeFL), a generalized framework that efficiently divides deep neural networks into submodels using both depthwise and widthwise scaling. To address the inconsistency arising from training multiple submodel architectures, NeFL decouples a subset of parameters from those being trained for each submodel. An averaging method is proposed to handle these decoupled parameters during aggregation. NeFL enables resource-constrained devices to effectively participate in the FL pipeline, facilitating larger datasets for model training. Experiments demonstrate that NeFL achieves performance gain, especially for the worst-case submodel compared to baseline approaches (7.63% improvement on CIFAR-100). Furthermore, NeFL aligns with recent advances in FL, such as leveraging pre-trained models and accounting for statistical heterogeneity.
24 July 2025 in IEEE Transactions on Mobile Computing
Federated learning (FL) is a distributed training paradigm that enables collaborative learning across clients without sharing local data, thereby preserving privacy. However, the increasing scale and complexity of modern deep models often exceed the computational or memory capabilities of edge devices. Furthermore, clients may be constrained to use heterogeneous model architectures due to hardware variability (e.g., ASICs, FPGAs) or proprietary requirements that prevent the disclosure or modification of local model structures. These practical considerations motivate the need for model-heterogeneous FL, where clients participate using distinct model architectures. In this work, we propose Generative Model-Aided Federated Learning (GeFL), a framework that enables cross-client knowledge sharing via a generative model trained in a federated manner. This generative model captures global data semantics and facilitates local training without requiring model homogeneity across clients. While GeFL achieves strong performance, empirical analysis reveals limitations in scalability and potential privacy leakage due to generative sample memorization. To address these concerns, we propose GeFL-F, which utilizes feature-level generative modeling. This approach enhances scalability to large client populations and mitigates privacy risks. Extensive experiments across image classification tasks demonstrate that both GeFL and GeFL-F offer competitive performance in heterogeneous settings.
29 July 2026 in Neurocomputing