Specific emitter identification (SEI) is a critical technology for wireless device security authentication. However, in practical applications, time-varying channel characteristics often obscure the radio frequency fingerprint features of emitters, significantly degrading recognition performance. Additionally, the high cost of sample labeling limits the effectiveness of incremental learning methods that rely on labeled data. To address this challenge, we propose a cross-domain semi-supervised IL approach for SEI (CDSIL-SEI). This method leverages domain adaptation to effectively utilize unlabeled samples, narrowing the feature gap between the source and target domains, and enabling incremental learning with limited labeled samples. Experimental results on the WiSig dataset show that CDSIL-SEI improves target-domain accuracy by about 15% over domain adaptation methods and achieves more than 10% higher source-domain accuracy than incremental learning methods, effectively alleviating catastrophic forgetting.

Autonomous navigation for UAV swarms in unknown environments holds critical importance for applications ranging from disaster response to military reconnaissance, yet faces multifaceted challenges including dynamic obstacle avoidance, real-time computational constraints, and absence of environmental priors. Conventional path planning methods (e.g., artificial potential fields, APF) frequently succumb to local optima under global information scarcity, while deep reinforcement learning (DRL) approaches, despite their environmental adaptability, suffer from low sample efficiency and slow convergence. This study proposes a Direct Policy Transfer-enhanced Multi-Agent Proximal Policy Optimization framework (DPT-MAPPO), merging APF spatial modeling efficiency with DRL adaptive optimization. Specially, the framework accelerates environment-task mapping during early training via APF-guided policy priors and resolves inherent local minima in conventional APF through DRL-driven stochastic exploration. Experimental validation across three representative unknown scenarios demonstrates that the proposed method attained 4-fold acceleration in convergence compared to MAPPO algorithms and 16% higher success rate compared to APF algorithms, confirming the framework’s dual advantages in safe navigation and computational efficiency for complex unknown environments.

In addressing the path planning problem, recent works consider the integration of the traditional A* algorithm with deep reinforcement learning, employing artificial neural networks as heuristic function, and training the neural networks with reinforcement learning, has been demonstrated to enhance the efficiency of path search in complex environments. However, two problems exist. 1) network retraining is required when encountering unfamiliar environments, 2) following the framework of the A* algorithm leads to a limited improvement in search efficiency. To address these problems, this paper proposes a hierarchical path planning framework. The upper level planning transforms the global path planning problem into a multi-stage path planning problem for a sequence of subgoal points according to the shortest distance principle, which leads to the improvement of the planning efficiency. The lower level planning implements path planning to the subgoal points by proposing a Local-environment Reinforcement Learning A* (LRLA*) mechanism, in which a heuristic function neural network with local environment as input is constructed and trained by self-generated paths, so that the generalization of the algorithm is guaranteed. Based on the hierarchical framework, a Hierarchical Reinforcement Learning A* (HRLA*) path planning algorithm is designed. The experimental results demonstrate that the HRLA* algorithm can rapidly generate a bounded suboptimal path in randomly generated environments after one-shot training, eliminating the need for retraining. Furthermore, the number of node expansions is reduced by more than 80% compared to the state-of-the-art algorithm, with the path cost loss comparable to existing algorithms.

Unmanned aerial vehicles (UAVs)-based cooperative target search plays an important role in the scenario of complex and urgent search missions. However, the limited battery life and onboard computational capacity make it challenging for UAVs to accomplish the search mission effectively. This paper investigates the problem of joint computation offloading and trajectory planning for a multi-UAV system integrated with edge computing. Our goal is to minimize the total uncertainty of the target search area while considering the limited energy and search time of UAVs. To adapt to highly dynamic environments, a deep reinforcement learning (DRL) framework is applied in this paper. We propose a novel approach of Action Masks and Branch-network based-Proximal Policy Optimization (AMB-PPO), by which a multi-branch network structure and action masks are introduced into PPO to address the challenges of large joint action spaces and low learning efficiency. The numerical results demonstrate that, when compared to other existing DRL methods, the proposed AMB-PPO algorithm performs better in reducing the total uncertainty of the target search area with higher learning efficiency and faster algorithm convergence.

Unmanned aerial vehicles (UAVs), also known as drones, have already been widely used in wireless networks. UAV-assisted integrated communication and sensing (ISAC) networks are feasible solutions to many challenging scenarios in which ground users (GUs) are inaccessible by terrestrial networks. However, the uncertainty of GU's locations undermines the performance of UAV-assisted networks, especially for UAV trajectory designs. To tackle this issue, we formulate this optimal UAV trajectory design problem to a catenary shape determination problem, which transforms the objective of maximizing the overall performance to that of minimizing the potential of the catenary. In the proposed scheme, an arbitrary partial distribution of GU's locations is represented by a matter with areal mass density in an artificial potential field (APF). To obtain an optimal solution, we derive a second-order mechanical equation representing the shape of this catenary, by analyzing its static equilibrium state when achieving minimal potential. Different from conventional path discretization methods, the obtained trajectory solution in this paper is a continuous-form second-order equation with remarkable path compression. The numerical results show that, when compared to conventional UAV trajectory optimization methods, the proposed approach can achieve an optimal solution of UAV trajectory with low computational complexity. It further demonstrates that the proposed approach can flexibly and continuously adjust the UAV trajectory under the scenario of uncertain GU's locations.

Elliptic positioning system offers a precise alternative to global navigation satellite system (GNSS). However, ranging measurements upon a single UAV only delineate the location estimates to a spherical region. Data infusion from inertial measurement units (IMUs) may refine these estimates, while its fidelity is undermined by IMUs’ lack of self alignment to a specified reference frame. In this paper, we explore the minimal number of assisted UAVs or anchors for absolute positioning, i.e., location and alignment in a fixed frame, during complete GNSS outages. We first prove that the observability establishes under a UAV in a 3-D trajectory or a pair of static anchors. The two numbers are new theoretical limit, significantly lower than the three anchors in 2-D or four in 3-D scenarios required by traditional theorems. We propose a sequential scheme for the multi-parameter estimation problem ensuring rapid convergence. An iterative solution is derived, flexible to UAV and anchor-based configurations, that provides instant location updates free of computational overhead. Thereafter, we also circumvent NLOS effects by employing inverse estimation of range. Accordingly, we devise a tiered positioning framework that commences with a location-unknown UAV to first cooperate with LOS anchors, and then extend the service to UE via a single NLOS link outside anchors’ coverage. In the experiments, the proposed scheme reaches (10−2)° orientation alignment and centimeter-level accuracy in NLOS scenarios, which attains the Cramer-Rao lower bound (CRLB) accuracy. Moreover, the accuracy notably exceeds the noise level of ranging measurements at high sampling rate, and also shows robustness against local clock drifting.

In this paper, we consider a unmanned aerial vehicle (UAV) aided integrated sensing and communications (ISAC) network with moving ground users in constant-velocity trajectory. A global optimal trajectory design scheme is proposed including a continuous analytic solution as well as optimization condition, which is theoretically different from numerical schemes obtaining a discrete piece-wise solution with approximate optimality. However, it is challenging to maximize the performance over entire infinite time slots in moving-user scenarios. By projecting the trajectory onto a user-relative coordinate frame, we reduce the performance to a location-determined function, which is in physical equivalence to an artificial potential field (APF). Accordingly, the optimization problem is reformulated to the shape determination problem of a density-varying catenary in the APF. By performing force analysis, we describe the topology of the catenary via a second-order differential equation determined by a boundary, i.e., any three combination of the location, orientation and turning curvature at arbitrary waypoints. Equivalently, the representation of the continuous solution is minimized in ultra-low-dimension parameter space and offers a flexible and lightning-speed design practice. Through complexity analysis, we find that the complexity of the proposed analytic scheme is significantly lower than the traditional discrete schemes. Further, we also prove that the global optimality, existence and uniqueness of the solution holds under a condition of a strong applicability to general sensing and communications (S&C) services. In simulation, the equivalence is confirmed and the results show global optimality, low-complexity and the high flexibility under avoidance, crossing and G-force limit.

Knowledge graph embedding, representing entities and relations in the knowledge graphs with high-dimensional vectors, has made significant progress in link prediction.More researchers have explored the representational capabilities of models in recent years.That is, they investigate better representational models to fit symmetry/antisymmetry and combination relationships.The current embedding models are more inclined to utilize the identical vector for the same entity in various triples to measure the matching performance.The observation that measuring the rationality of specific triples means comparing the matching degree of the specific attributes associated with the relations is well-known.Inspired by this fact, this paper designs Semantic Filter Based on Relations(SFBR) to extract the required attributes of the entities.Then the rationality of triples is compared under these extracted attributes through the traditional embedding models.The semantic filter module can be added to most geometric and tensor decomposition models with minimal additional memory.Experiments on the benchmark datasets show that the semantic filter based on relations can suppress the impact of other attribute dimensions and improve link prediction performance.The tensor decomposition models with SFBR have achieved state-of-the-art.

Time delay estimation (TDE) is of great importance in reconnaissance and passive localization. Though TDE could leverage compressive sensing (CS) kernel to enjoy the advantage of sub-Nyquist samples, the estimation accuracy and robustness are inclined to be reduced by jamming or noise. To address the issue, this letter proposes a scheme of defensive compressive time delay estimation (DCTDE) with a novel sensing kernel optimization method and a defense augmentation strategy for the kernel. Inspired by the information bottleneck (IB) theory, we deduce a new analytic principle to enhance the estimation capability of the kernel, which focuses on the fidelity to the non-compressive scheme and the relevant information provided for TDE task.The strategy adopts an adversarial idea for the anti-spoofing design of kernel. Via simulations, the proposed scheme shows promising estimation accuracy, robustness, and high adaptability in both Gaussian noise and jamming environments, and more competitiveness than the Nyquist scheme.

Pulse position modulation‐ultra wideband (PPM–UWB) communication signal is hard to detect and sample directly, owing to its ultra‐low power spectral density and wide bandwidth. There are already some researches on using analogue‐to‐information converter (AIC) technology and compressed sensing (CS) theory to under‐sample and detect PPM‐UWB communication signal, utilising its sparseness in time domain. However, greedy algorithm lacks of restriction on sparseness of reconstructed vector, while common restrictions on sparseness (e.g. convex optimisation) has high computational complexity. To solve these problems, a combinatorial optimisation method is proposed in this study to detect PPM–UWB communication signal based on CS and AIC. Reconstruction error and sparseness of reconstructed vector are restricted by l 2 ‐ and l p ‐norms, respectively. l p ‐norm (0 < p < 1), which is a non‐convex function, has stricter restriction on sparseness than l 1 ‐norm. Meanwhile, the steepest descent method is adopted for l p ‐norm optimisation, which can rapidly converge to objective values. Proposed method has more comprehensive restriction than greedy algorithm and convex optimisation, while maintain low complexity in computation as greedy algorithm. Numerical experiments demonstrate the validity of proposed method.

Pulse position modulation‐ultra wideband (PPM‐UWB) communication adopts ultra narrow pulse as the transmitted signal. Owing to the low‐power spectral density and ultra wide bandwidth, it is difficult to detect and sample PPM‐UWB signal directly. There are already some researches on using compressed sensing (CS) theory for UWB communication with lower sampling speed. However, these methods take the sparseness of pulse position or transmission channel into account separately and they are unfit for practical communication. To solve these problems, a dual‐sparse reconstruction method is proposed in this paper to process PPM‐UWB communication signal based on CS theory. Proposed method designs the target signal which needed to be reconstructed as a dual‐sparse vector. This vector combines the sparseness of PPM pulse position and UWB channel multi‐paths simultaneously, hence it has dual sparseness. The information code can be demodulated from reconstructed dual‐sparse vector directly by using energy detection method. Extensive numerical simulations demonstrate the validity and applicability of proposed method.