The widespread adoption of Unmanned Aerial Vehicles (UAVs) or drones has introduced significant security and privacy challenges, particularly concerning unauthorized drone activities near sensitive areas. To address these concerns, we propose a novel drone detection model, WRN-YOLO, which integrates the Wide Residual Network (WRN) architecture with the You Only Look Once (YOLO) object detection framework. This integration enhances feature extraction capabilities, leading to improved detection accuracy. Through comprehensive ablation studies, we have identified the optimal YOLO variant that synergizes with our backbone modifications, ensuring superior performance in diverse scenarios. Recognizing the complexities of real-world environments, we have also developed a synthetic dataset designed to train our WRN-YOLO. This dataset encompasses a variety of challenging conditions, including intricate backgrounds and the presence of confounding elements, to robustly assess the model's efficacy. Experimental results demonstrate that our method significantly outperforms existing models in accurately detecting drones amidst complex scenes, offering a promising solution for real-time UAV threat mitigation. The proposed approach ranked Top 3 in the 8th WOSDETC Drone-vs-Bird Detection Challenge. Our source code and synthetic dataset are publicly available at https://github.com/yjwong1999/IJCNN2025-DvB.

The use of aerial base stations (ABSs) has gained significant interest due to their deployment flexibility for coverage and capacity enhancements. However, the load balancing problem in multi-ABS networks, which is crucial for quality of service (QoS) provisioning and fair ABS utilization, remains largely unresolved. This paper investigates the load balancing problem for multi-ABS networks with consideration of backhaul limitations, while taking into account the delay-QoS performance of the networks. Firstly, we quantify the load of each ABS with the effective capacity (EC), which is characterized by a latency parameter known as the delay-QoS exponent. Then, we design a load balancing utility function based on the generalized α-fairness concept, which provides the flexibility to the network to achieve various degrees of load balance among ABSs. Next, a joint ABS placement and user association problem is formulated with the aim to maximize the load balancing utility function, with consideration of backhaul capacity constraints, physical isolation between ABSs and ABSs’ limited user capacity. Since the problem is a mixed-integer programming problem which is generally difficult to solve optimally, we develop an efficient solution framework, which optimizes the 3D placements of ABSs using swarm intelligence and user association via convex optimization. Simulation results show that our proposed scheme outperforms existing schemes in terms of load balance, probability of blocking, and EC.

Recently, Deep Reinforcement Learning (DRL) has been applied to enhance the Quality of Experience (QoE) of Adaptive Bitrate Streaming (ABR) by adjusting the video quality level in real time based on instantaneous network conditions. To build a state-of-the-art DRL-based ABR (DRLABR) algorithm, it must learn from the clients’ actual network and video streaming behavior. However, collecting such data directly from clients introduces several challenges, including privacy concerns, high bandwidth consumption, and the straggler effect—where poor network conditions of certain clients delay the training process, as DRLABR’s performance is highly dependent on network interactions. To overcome these limitations, we propose a decentralized training approach for DRLABR using a Federated Learning (FL) framework. Instead of gathering raw data, clients train their local DRLABR models independently and send only model updates to the central server. To address the straggler issue, we propose to desynchronize the FL update rules, allowing clients to contribute their model updates at their own pace, regardless of varying network conditions. In addition, we design a DRL-based client selection mechanism to prevent oversampling of high-bandwidth clients, which could lead to model divergence, thereby ensuring balanced participation and improving the overall training efficiency. We validate our approach through a comprehensive simulation encompassing diverse video content and real-world network traces, simulating a wide range of streaming activities. Our results show that the proposed framework significantly outperforms conventional FedAvg and FedAsync methods, achieving the highest average QoE score of 2.02 and reducing the total training latency by 21.26%.

Deep learning has significantly advanced the field of building extraction from remote sensing images, providing robust solutions for identifying and delineating building footprints. However, a major challenge persists in the form of domain adaptation, particularly when addressing cross-city variations. The primary challenge lies in the significant differences in building appearances across cities, influenced by variations in building shapes and environmental characteristics. Consequently, models trained on data from one city often struggle to accurately identify buildings in another city. In this paper, we address this challenge from a data-centric perspective, focusing on diversifying the training set. Our empirical results show that improving data diversity via open-source datasets and diffusion augmentation significantly improved the performance of the segmentation model. Our baseline model, trained with no extra dataset, only achieved a private F1 score of 0.663. On the other hand, our model trained with the additional Las Vegas building footprints extracted from the Microsoft Building Footprint dataset, achieved a high private F1 score of 0.703. Surprisingly, we found that diffusion augmentation helps improve our model score to 0.681 without requiring an extra dataset, which is higher than the baseline model. Finally, we also experimented with the Non-Maximal Suppression (NMS) hyperparameter to improve the model’s performance in segmenting dense and small objects, which gave us a high private F1 score of 0.897. These techniques ultimately led our solution to rank 1st in the competition. Our source code and the pretrained models are publicly available at https://github.com/DoubleY-BEGC2024/OurSolution.

The continuous evolution of cellular networks has resulted in the rapid increase in both mobile applications and devices in the Internet of Vehicles. The introduction of the multi-access edge computing method makes it possible for vehicles in remote areas to offload their computational tasks, which can effectively relieve the computing pressure of local devices and reduce the computational delay as well. Tasks offloading for multi-user is a resource competition problem, especially in dynamic environments, which is difficult to be solved by traditional algorithms. In this article, we propose a two-layer hybrid system with local and edge computing, providing convenient computing and offloading services for vehicle users in dual dynamic scenarios of task generation and vehicle mobility. The delay and queuing situations are considered comprehensively in the formulated optimization problem, which can be solved by the proposed deep deterministic policy gradient-based computation offloading algorithm. The offloading process of the vehicle tasks in dynamic scenarios is transformed into a Markov decision process to obtain the offloading strategy. Simulation results demonstrate the performance advantages of two-tier computing architecture. Compared with random offloading, deep Q network-based offloading, and local computing, the algorithm proposed in this article gains the highest average reward of tasks. Besides that, numerical results also prove that our algorithm has the lowest average delay under different computing capabilities of edge servers.

Backscatter communications (BC) have recently emerged as a promising technology for sixth-generation (6G) networks with device-to-device (D2D) communications to support energy-efficient Internet of Things communications. This paper investigates the power allocation problem for energy-efficient 6G networks with backscatter-enabled D2D (BC-D2D) communications. To this end, we formulate the power allocation problem as a biobjective optimization problem that jointly maximizes the sum data rate and minimizes the power consumption of the networks, subject to energy-harvesting, reflection coefficient (RC) and power constraints. To solve this problem, suboptimal RCs are first analytically obtained based on the energy-harvesting constraints of the BC-D2D transmitters. Next, the biobjective optimization problem is transformed into a convex, single-objective power allocation optimization problem using the weighted sum approach, which is then solved using convex optimization. Results show that the proposed scheme outperforms the baseline schemes in energy efficiency under scenarios with different numbers of cellular user equipment and BC-D2D pairs.

This paper investigates the performance of non-orthogonal multiple access (NOMA) in vehicular networks where a base station (BS) communicates with the vehicles moving away from the BS with single-input multiple-output. To combine the signals received at the antennas, diversity combining techniques such as maximal ratio combining (MRC) and selection combining (SC) are performed at the receiver of each vehicle. However, in practice, the expected performance from the diversity techniques may not be achieved due to the fact that all the diversity branches are not independent and identically distributed (i.i.d) all the time. In this context, analytical expressions of the outage probability and ergodic sum rate are derived for the considered vehicular networks with the assumption of independent but not necessarily identically distributed (i.n.i.d) Nakagami-${m}$fading channels. The performance analysis of NOMA vehicular networks is also extended for multiple-input multiple-output antenna configurations and evaluated in the presence of successive interference cancellation (SIC) error propagation. The obtained analytical results are validated by Monte Carlo simulations. Furthermore, the performance of NOMA is verified with conventional orthogonal multiple access (OMA) for fading parameter$m=1$and$m=2$with perfect channel knowledge and channel estimation. Numerical results show that NOMA outperforms the conventional OMA by approximately 20% and has high sum rate with i.n.i.d as well as i.i.d channel consideration. However, i.n.i.d consideration degrades the performance of NOMA and OMA as the diversity gain achieved with i.n.i.d consideration is less as compared to i.i.d consideration. The performance is further deteriorated with SIC error and channel estimation.

In limited space scenarios, the antennas in the multi-antenna cognitive radio (CR) system are closely spaced and often experience correlation among them. In this paper, the sensing performance of arbitrary correlated antennas over Nakagami-m fading channel for the mobile CR user is analysed. In particular, the analytical expression for the average detection probability for a mobile CR user employing the selection combining with triple arbitrary correlated diversity branches is derived as a special case. Furthermore, to characterize the performance of energy detector under mobility, the area under the curve of receiver operating characteristic is analysed. The derived expressions converge quickly due to the monotonically decreasing hypergeometric function of two variables. The Monte Carlo simulations substantiate the analytical expressions. Results indicate that antenna correlation deteriorates detection performance. Moreover, the high speed of CR users further decreases the detection performance, especially in the deep fading channel scenarios. This work provides a realistic sensing framework for the CR enabled vehicles.

In this work, energy detection‐based spectrum sensing for multiple antenna receiver under the effect of mobility is investigated by considering L number of correlated antenna branches. The authors consider the uniform, exponential and arbitrarily correlation among the antenna branches based on the spacing between them. The moment generating function (MGF) approach is applied to obtain the statistical knowledge of the received signal to noise ratio because the Laplace domain behaviour will help to derive the closed‐form expressions using simple algebraic operations. They derived the closed‐form expressions for the detection probability over Nakagami‐ m fading, in terms of Lauricella and Confluent Hypergeometric function for maximal ratio combining (MRC) and equal gain combining (EGC) diversity techniques under the effect of vehicle mobility. Monte‐Carlo simulation is carried out to validate the derived analytical expressions. The results show that the degradation in detection performance due to fading correlation can be reduced by choosing the appropriate diversity scheme and by increasing the number of antennas. Furthermore, they also found that at high fading parameter ( ) value, the low value of the probability of false alarm and highly correlated fading, MRC works better than EGC for high relative velocity.

Dynamic Spectrum Access (DSA)/Cognitive Radio (CR) systems access the channel in an opportunistic, noninterfering manner with the primary network, thus being a promising approach to solve the problem of spectrum scarcity. Energy Detection, a spectrum sensing technique for DSA/CR systems, is widely used for blind sensing of unused frequency bands due to its non-parametric sensing ability and computationally low complexity. However, spectrum sensing becomes more challenging in Cognitive Vehicular Networks (CVNs) due to Secondary User's (SU's) mobility and often yields a detection performance loss as compared to static scenarios. In order to mitigate the impact of reduced detection performance due to mobility, the usage of an improved version of energy detection technique is proposed in this paper. Usage of Improved Energy Detection (IED) technique in CVNs results more than 10% increment in DSA/CR system performance. In this paper, we study the joint impact of SU's sensing range, PU's protection range and SU's mobility model on the PU Activity using IED technique in CVNs, with detection probability and probability of false alarm as the performance metrics. Also, we derive a closed form expression for the probability of PU being inside SU's sensing range. Based on the proposed framework, numerical results show great agreement with analysis, yielding a superior performance.

Transmission of compressed video over wireless channels remains a challenging task due to the inherent high bit error rate (BER) and channel quality fluctuation of a typical wireless channel. An improved wireless video transmission scheme is proposed in this paper. First, an enhanced video error propagation model,namely expected number of macro blocks error propagation (ENMEP), is proposed. A new unequal error protection (UEP) scheme which takes into consideration the non-uniformly distributed importance of frames in a group of pictures (GOP) and macro blocks in a video frame is also proposed in this paper. Finally, an efficient forward error correction (FEC) allocation algorithm for our UEP scheme is proposed by making use of our ENMEP video error propagation model. Simulation results show that our proposed UEP scheme using efficient FEC allocation algorithm outperforms the classical equal error protection (EEP) scheme and also the previous UEP scheme.

This paper proposes new divergence detection techniques for implementation within in-service non-intrusive measurement devices (INMDs) in public switched telephone networks (PSTNs). The in-service non-intrusive measurement system of interest is used to monitor the delivered quality of speech (QoS) by monitoring the echoes in the telephony network. INMDs are usually based on a class of least mean square (LMS) digital adaptive filters (DAFs). The performance criterion is defined by the modelling convergence rate derived from the optimal Wiener weights, and the excitation for the DAFs is conversational speech. Four types of divergence detectors (DD) are proposed. These are energy divergence detectors (EDD), log energy divergence detectors (LDD), zero crossing divergence detectors (ZDD) and autocorrelation coefficient divergence detectors (ADD). The proposed DDs are based on the detection of voiced/unvoiced/silence periods and as such act as pattern classifiers. Experimental observations have shown that divergence occurs during the low energy unvoiced segments in high-noise environments. The tap-weight coefficients of the DAF are updated with the new value during the voiced segment while the update of the tap-weight coefficients during unvoiced segments of the speech is frozen. This result is then compared with the perfect divergence detector, which employs the Wiener weight theory. The DD techniques reported produce a significant improvement in the system's performance in a noise-impaired environment. Over one second adaptation (8000 samples) the energy divergence detector, the log energy divergence detector, the autocorrelation divergence detector and the zero crossing divergence detector gave model improvements of 16.93 dB, 15.81 db, 12.48 dB and 11.62 dB respectively at echo to noise ratio (e/N) of 0 dB. The proposed DDs compare well with the ideal (nonimplementable) Wiener DD which gives an improvement of 20.92 dB.