Browse Topic: Unmanned aerial vehicles

Items (1,628)
Polymer Electrolyte Membrane (PEM) Fuel Cell driven vehicles are an alternate mode of transportation systems, most suitable for long-haul heavy-duty transportation and being explored for usage in Unmanned Aerial Vehicles (UAV). Hydrogen reacts on the anode side, while air reacts on the cathode side of the crucial Membrane Electrode Assembly (MEA) to produce electricity primarily, and water and heat as byproducts. The performance of the fuel cell is assessed by its voltage vs current characteristic (polarization) curve, which is affected by the design and operating parameters such as the design of MEA, Gas Diffusion Layer (GDL), gas channels, cooling system, and operating gas flow rates, pressure, temperature and humidity. In this study, the effect of some of these design and operating parameters on polarization curve is investigated using MATLAB Simulink’s 1D model. Using this 1D model, optimal designs can be quickly obtained when compared to 3D CFD simulations. Polarization curves are obtained using the 1D model and comparison of the fuel cell performance is made with respect to different parameters such as MEA and GDL thickness, gas channel width, and stack temperature etc. The obtained optimum stack temperature from this study help to maintain proper hydration level of membrane for the obtained optimum configuration which is crucial to improve fuel cell vehicle performance.
Karlapalem, Vidyadhar
When operating at low altitude, UAVs are exposed to gusts, unsteady aerodynamic loads, and electromagnetic (EM) interference at the same time. A fixed-geometry airframe has no recourse against this kind of compound threat. In this paper, a four-stage methodology called Disturbance–Response–Mapping–Optimization (DRMO) is put forward. It follows the propagation path of external disturbances — from macro scale down to micro scale — until the disturbances surface as measurable performance loss. Three groups of variables enter the formulation: morphing configuration (span reduction ratio r, camber adjustment), skin topology (ZPR honeycomb, auxetic, corrugated), and functional compensation. Calculations are carried out on a 2.0 m span, 5 kg platform. Helmbold lift-slope theory combined with a simplified flutter model shows that r = 0.15–0.20 cuts gust load by 17–23 % and keeps the flutter speed ratio above 0.95. Skin strain-to-frequency drift is captured by a power law (k = 1.8, n = 1.6) whose coefficients originate from published silicone-substrate antenna data. A ±15 % perturbation of these coefficients does not move the design-window edges outside the acceptable band. The intersection of the three constraint boundaries fixes a passive window at r = 0.15–0.20. Adding piezoelectric LQR stiffness control widens it to r = 0.25 without degrading flutter safety.
Wang, Lechen, Zhou, Tingyao, Qu, Fangqi, Zheng, Jiaou, Chu, Lingling, Jia, Yilin, Yan, Jingjing
To regulate low-altitude UAV operations, this paper proposes a method for constructing a low-altitude layered route network. First, a visibility-graph-based obstacle environment modeling method is adopted to transform geometric constraints in complex low-altitude space into a graph representation, and candidate flyable networks are generated at different altitude layers. On this basis, a set of primary paths and low-overlap alternative paths is generated for multiple origin-destination pairs. To address the redundancy and discreteness of the resulting path set, multiple OD paths are combined into an initial low-altitude airspace route network through path screening, spatial aggregation, and shared-structure extraction. Furthermore, a minimum spanning tree is used to extract the network backbone for topological simplification. On this basis, redundant connections and reinforced critical route segments are introduced through network augmentation and optimization strategies, thereby constructing a low-altitude airspace network that integrates topological connectivity, structural compactness, operational efficiency, and redundancy robustness. Using Longhua District of Shenzhen as the case study area, this paper carries out simulation experiments on airspace planning. The results indicate that the proposed method requires only a marginal expansion of network size, yet it can markedly lower the average detour level of key OD pairs, enhance the two-path reachability ratio, and preserve stronger OD service capability under both random and directional disturbances. Therefore, the method is able to support the efficient construction of layered route networks in low-altitude environments characterized by dense obstacles, while also providing a methodological basis for the operational management of low-altitude UAV systems.
Liu, Mingyu, Zhang, Ying, Yin, Jianan, Tian, Wen, Meng, Fanyu
Aiming at the complex dynamic environment with static obstacles and multiple UAVs, this paper proposes a path planning method for unmanned aerial vehicles (UAVs) based on the improved Artificial Potential Field (APF) method and Pigeon-Inspired Optimization (PIO). The own UAV generates a collision-free trajectory using the APF method with an obstacle repulsive potential field introduced, while other UAVs adopt PD control with an obstacle avoidance term superimposed for collision avoidance. Obstacle avoidance penalties are designed in the payoff functions of both sides, enabling the path planning to have both obstacle avoidance and cooperative deconfliction capabilities simultaneously. The PIO algorithm is introduced to optimize the potential field parameters of the own UAV and the control parameters of other UAVs respectively, making the strategies of both sides gradually approach the Nash equilibrium through iterative interaction. Simulation experiments are carried out in a 2D environment with static obstacles and multiple UAVs, and the results show that the proposed method can effectively guide the own UAV to avoid collisions and obstacles and reach the target point successfully. The PIO algorithm exhibits good convergence and high path quality in complex obstacle scenarios, verifying its effectiveness and superiority.
Xu, Lina, Liu, Liang, Yao, Yubo, Liu, Mengyan, Shi, Hua, Dai, Fei, Feng, Xiaozhou
In real-world operations, tasks such as urban inspection and forest search place extremely high demands on UAV autonomous obstacle avoidance and navigation. However, the classic DQN falls short in practical deployment due to its limitations in single-sensor input, insufficient exploration, and slow convergence. This paper introduces three key improvements to DQN: First, a dual-stream network is built to fuse visual and LiDAR information. Second, prioritized experience replay (PER) and NoisyNet are introduced to enhance sample efficiency and exploration quality, respectively. Third, a three-stage curriculum learning strategy is designed to gradually reduce the training difficulty. Experiments are conducted using TensorFlow 2.x and Gazebo, and the results show that our improved algorithm significantly outperforms DQN, DDQN, Dueling DQN, and PPO in terms of the success rate, path length, and adaptability to dynamic obstacle avoidance.
Lin, Shuai, Wen, Zhonghua, Lin, Xiong, Xia, Xueya
The vehicle-drone cooperative delivery mode has demonstrated significant potential in enhancing last-mile delivery efficiency by effectively integrating the transportation advantages of both vehicles and drones. However, increasingly complex last-mile delivery scenarios impose higher requirements on this cooperative approach. To accommodate the demand for routing flexibility arising from distribution network layouts featuring multi-depot collaboration and multi-type rendezvous points used for drone launch and retrieval, this paper explores the collaborative multi-depot vehicle routing problem (MDVRP) with drones, considering multi-type rendezvous points (CMDVRP-D-MTRP). Specifically, vehicles and drones can flexibly select their respective departure and return nodes for each distribution route, and drones can choose rendezvous points from either fixed parking nodes or temporary parking nodes converted from customer nodes that are actually visited by associated vehicles. First, based on the framework of the half-open multi-depot vehicle routing problem (HOMDVRP), the CMDVRP-D-MTRP is formulated as a Mixed-Integer Programming (MIP) model to minimize the total cost, including distribution and vehicle waiting costs. Second, by adapting benchmark instances of the MDVRP, this paper designs test instances tailored to the CMDVRP-D-MTRP covering various scales and distribution features. We apply the Cardinal Optimizer (COPT) solver to validate model feasibility and analyze the impact of problem scale on computational performance. Finally, based on the classification criteria of multi-depot collaboration and multi-type rendezvous points, numerical experiments are conducted across four delivery modes to verify the effectiveness of the proposed model. Compared to the traditional delivery mode that lacks multi-depot collaboration and relies solely on single-type rendezvous points, the proposed model for the CMDVRP-D-MTRP reduces the average total delivery cost by 20.58%.
Xiao, Shichang, Lu, Shengye, Yu, Shaohua
Water conservancy and transportation infrastructure are traditionally monitored using techniques including total stations, sensor networks, or global navigation satellite systems (GNSS). Although these techniques offer high accuracy at discrete points, they are unable to identify deformations outside of monitored locations, and increasing point density significantly increases costs. Growing interest has been shown in the recently developed unmanned aerial vehicle (UAV)–based remote sensing; nonetheless, most studies rely on single data sources, which results in poor accuracy and low robustness in complex situations. To overcome these drawbacks, this study proposes a deformation monitoring method based on UAV multisource data fusion. A high-precision fused model is created by combining high-resolution oblique imagery with airborne LiDAR point clouds. During aerial triangulation, preprocessed LiDAR point clouds serve as geometric constraints. This combines extensive texture information with high geometric accuracy to enable accurate registration within a single coordinate system. A two-stage registration strategy, comprising feature-based singular value decomposition (SVD) and the iterative closest point (ICP) algorithm, is employed for multi-period models. Surface displacements are measured using cloud-to-cloud (C2C) distances, which are then broken down into orthogonal X-, Y-, and Z-components to meet technical requirements. Validation on a reservoir dam shows that the proposed strategy greatly improves monitoring accuracy and reliability when compared to single-source approaches. The automated workflow provides a practical solution for long-term infrastructure monitoring by facilitating efficient data processing and the detection of subtle surface deformations.
Xu, Junhao, Ni, Guizhu, Li, Zheng, Hua, Yuansheng
Structural displacement is an important quantity in Structural Health Monitoring (SHM), as it directly reflects the condition and safety of structures. Vision-based measurement methods have been widely studied in recent years, mainly because they allow non-contact sensing and can cover relatively large areas. When these methods are applied to Unmanned Aerial Vehicle (UAV) platforms, however, additional challenges arise. In practical operation, platform motion and attitude changes tend to introduce drift into the measurements, which can noticeably affect the final results. In this study, a displacement measurement approach that combines vision sensing with a laser reference is investigated for both UAV-based and fixed-camera scenarios. A laser source installed on the ground projects a reference spot onto the structural surface, placed close to a multi-circle target. This setup provides a stable reference point that is less affected by platform movement. The displacement is then estimated by tracking the relative motion between the target and the laser spot, instead of relying directly on absolute measurements. With this design, a large part of the motion caused by the platform can be separated, allowing the structural displacement to be expressed in a ground-based coordinate system. The implementation includes several practical steps, such as adaptively selecting regions of interest, performing basic image processing, refining feature locations at the subpixel level, and updating scale factors during operation. Experiments with both UAV and stationary camera setups show that conventional measurements are strongly affected by low-frequency drift, while the laser-referenced results remain relatively stable across different measurement points. According to the Root Mean Square (RMS) analysis, about 77%–95% of the drift caused by platform motion can be reduced under UAV conditions. These results indicate that the method can serve as a practical option for displacement monitoring in UAV-based vision applications.
Li, Zheng, Xu, Junhao, Hua, Yuansheng
Manual highway inspection in conventional practice suffers from low operational efficiency, elevated safety risks for field personnel, and unavoidable inspection blind spots across key infrastructure sections. To address these prominent challenges, this research develops an integrated intelligent UAV inspection system that synergizes flight control technology, Beidou positioning, machine vision, and artificial intelligence algorithms. The system enables rapid detection and accurate identification of structural defects on highway pavements, bridges, slopes, and other critical infrastructure components. We conduct targeted optimizations on both the YOLOv7-RDD object detection algorithm and the DeepLabV3+ segmentation algorithm, optimizations that markedly boost the accuracy and environmental robustness of defect recognition in complex highway scenarios. A dedicated deployment scheme featuring automatic UAV airports and intelligent route planning is also proposed to underpin fully automated, high-frequency inspection operations. Field engineering applications have validated the system’s practical effectiveness: defect recognition accuracy surpasses 95% in real-world highway environments, while overall inspection efficiency is improved by approximately 80%. This research delivers pivotal technical support for the intelligent maintenance and refined management of transportation infrastructure networks.
Bai, Xing, Xie, Yaoxuan
Unmanned aerial vehicles (UAVs) equipped with chemical contaminant sensors offer an innovative approach for rapid and safe remote chemical reconnaissance. However, the proliferation of diverse UAV platforms presents a critical challenge: how to select the optimal platform for a given mission scenario. This paper establishes a systematic suitability analysis framework for deploying chemical contaminant sensors on three mainstream UAV platforms-multirotor, fixed-wing, and vertical take-off and landing (VTOL) types. We analyze payload capacity, endurance, flight stability, environmental adaptability, and mission system integration. To reduce decision subjectivity, we develop an Analytic Hierarchy Process (AHP)-based quantitative evaluation model incorporating expert surveys and consistency tests to validate the weighting process. Beyond the evaluation framework, we explore key engineering challenges in platform-sensor integration, including aerodynamic interference, vibration and electromagnetic compatibility, and power management. We compare quantitative measurements and qualitative features across platforms in simulated mission scenarios. The resulting selection criteria provide a theoretical basis and a reproducible decision-support framework for developing chemical detection systems applicable to emergency response, environmental monitoring, and public safety.
Liang, Ting, Wen, Hao, Qi, Yelin, Yang, Wen
Complex terrain imposes highly nonlinear constraints on multi-UAV cooperative path planning. It also leads to path redundancy and inefficient information sharing among UAVs. To address these issues, this paper proposes a cooperative path-planning method based on an improved Dung Beetle Optimizer (DBO) under dynamic constraints. First, a multi-objective path optimization model is constructed. The model considers detection efficiency, path smoothness, obstacle-avoidance cost, and energy consumption. State-space equations are introduced to constrain the path generation process. This ensures that the planned paths satisfy UAV dynamic constraints and remain feasible for control implementation. Second, to reduce the risk of premature convergence in the standard DBO, the Latin hypercube sampling-Orchestrated Dung Beetle Optimizer (LO-DBO) is proposed. Latin hypercube sampling is used to generate a uniformly distributed initial population in the solution space, thereby enhancing the diversity of the initial population. A dynamic convergence factor is then introduced to balance global exploration in the early stage and local exploitation in the later stage. In addition, a cooperative search mechanism is designed to help individuals escape local optima and improve the algorithm’s global search capability. Experiments on CEC2017 benchmark functions and multi-UAV cooperative path-planning scenarios show that LO-DBO achieves higher convergence accuracy, better solution quality, and greater planning efficiency than several representative optimization algorithms. Compared with other improved DBO, LO-DBO can coordinate multi-UAV resources more effectively. It also enables safe obstacle avoidance and efficient full-coverage inspection operations in complex terrain scenarios.
Wu, Binglin, Song, Yifei, Zhu, Chunhua, Li, Ning
With the rapid development of civil unmanned aircraft systems (UAS), their operating scenarios are becoming more and more complex. The operation mode of one pilot controlling one or more unmanned aircraft has been unable to meet the needs of specific scenarios, such as logistics scenarios. Distributed operation has gradually become the future development trend of UAS in these scenarios. This paper expounds the theoretical basis, structural framework, and implementation path of the grading of the distributed operation level. In this paper, the evaluation framework is established by using the two dimensions of system automation level (AL) and safety assurance (SA) when grading the operation level (OL), and the evaluation of the operation level is supported by standardized test methods. The establishment of the method in this paper can fill the gap in the operation and management of drones, and provide a basis for operators to apply for operating licenses and for government supervision.
Ma, Mengdi, Tian, Lingling
The deep integration of edge computing, blockchain, and UAV technology with the low-altitude economy has rapidly expanded the scale and scope of low-altitude communication networks. To support the rapid development of the low-altitude economy, the number of heterogeneous terminal devices is growing exponentially, posing new challenges to data transmission security within low-altitude data networks. Most UAVs and terminal devices adopt lightweight designs to maintain flight weight, resulting in limited computational and storage resources that cannot support mature cryptographic schemes. This vulnerability makes data collected by such devices susceptible to eavesdropping and tampering. Furthermore, standard symmetric encryption algorithms fail to decouple the spatio-temporal correlation of data within low-altitude networks, creating security risks of interception or modification in open low-altitude network environments. This paper designs and implements a highly secure, lightweight, low-altitude data encryption protection scheme based on the SM4 cryptographic algorithm and chaos theory. A hardware-software co-processing encryption system was developed on the ZYNQ platform. This solution introduces chaotic mappings into the SM4 key expansion phase in the software layer to enhance network security, while a cyclic reuse architecture and dual-buffer structure achieve low resource overhead algorithm acceleration on the programmable hardware layer, ensuring data throughput. Verification results from the FPGA-based test platform demonstrate that the proposed system exhibits effectiveness and efficiency under low resource overhead conditions. It can be deployed as an endpoint or additional module within low-altitude networks to safeguard data security.
Qin, Zhikang, Yu, Chengxiao, Chen, Xin, Liu, Ying
Asynchronous federated learning (AFL) is essential for low-altitude networks due to its benefits in reducing communication bandwidth, lowering transmission delays, and preserving data privacy. However, the highly dynamic mobility of drones, combined with inherent heterogeneities in AFL, including variations in data volume, data distribution, and device capability, introduces severe challenges, such as inconsistent model update frequencies, biased local training, and inefficient global aggregation. Consequently, these issues lead to significant accuracy degradation and hinder practical deployment. In this paper, we propose a novel AFL framework tailored for low-altitude networks to address these challenges, which combines a multi-dimensional clustering mechanism with a multi-weight aggregation mechanism (MDC-MWA). To mitigate local model bias induced by data heterogeneity, the multi-dimensional clustering allows the cloud server to compute a comprehensive score for each drone based on multiple dimensions, including the label distribution, the gradient direction, and the local data proportion. It then employs K-Means clustering to partition drones into highly homogeneous subgroups, each managed by a dedicated roadside unit (RSU). Within each subgroup, after performing local updates, each drone synchronously uploads parameters to its RSU. Between these subgroups, to address inconsistent update frequencies caused by imbalanced computing capabilities, RSUs perform asynchronous training via the multi-weight aggregation mechanism, which updates the global model considering staleness weight and direction consistency weight. Experimental results demonstrate that the proposed method outperforms baseline approaches in highly heterogeneous scenarios, achieving a 4.74% absolute improvement in training accuracy on the CIFAR-10 dataset. This work provides a practical framework for distributed intelligent applications in dynamic and resource-constrained low-altitude networks.
Cai, Jiawei, Cui, Wei, Wen, Ying, Chen, Zhentao, Wu, Maoqiang
In traditional multi-unmanned aerial vehicles (UAVs) cooperative strategies within complex irregular regions, the low coverage efficiency, boundary constraint failure, and insufficient robustness bring the challenges. This paper proposes a cooperative search-andattack algorithm, which integrates adaptive mission area discretization and distributed intelligent particle swarm optimization (DIPSO). An adaptive grid model is constructed with ray casting algorithm to discretize arbitrary polygonal mission areas precisely. It can effectively reduce the invalid searches in non-mission zones by abandoning the traditional rectangular assumption. Moreover, a multi-objective fitness function (MOFF) is designed by incorporating coverage ratio, path length, collision penalty, and load balancing. With the discrete path manipulation operators (PMOs) and dynamic learning factors, the proposed method achieves distributed trajectory planning in convex polygonal spaces. Simulation results demonstrate that compared with counterparts, the proposed method performs closer boundaries to actual ones. Furthermore, it achieves 100% coverage while shorten the mission time significantly, validating its adaptability in various convex polygonal scenarios.
Zhou, Yunbo, Zhang, Di, Xia, Wei
Unmanned Aerial Vehicles (UAVs) are essential for traffic monitoring in Intelligent Transportation Systems (ITS). However, the detection of traffic objects in drone-view imagery still faces serious challenges due to drastic scale variations, dense small objects, and complex background clutter. Existing detectors often struggle to effectively separate targets from background noise and lack efficient cross-scale feature interaction mechanisms. This paper introduces ED-DETR, a refined real-time Detection Transformer specifically designed to optimize performance in UAV scenarios. Specifically, a Dual-Stream Coordinate Attention (DCA) module is designed, which can effectively extract significant texture features while preserving context information through parallel maximum pooling and average pooling operations. In addition, this paper introduces an efficient multi-scale attention structure in the feature fusion neck, which combines grouping convolution and cross-spatial learning to significantly improve the positioning ability of small targets. A large number of experiments on the VisDrone2019 dataset show that ED-DETR achieves an mAP of 40.4%, which is significantly higher than that achieved by RT-DETR and other state-of-the-art detectors. By harmonizing high detection accuracy with low computational overhead, the proposed framework enables effective real-time monitoring of aerial traffic scenarios.
Jiang, Haiming, Zhu, Liang, Liu, Jianhua
In recent years, Reinforcement Learning (RL) has been increasingly applied to the collision avoidance systems of Unmanned Aerial Vehicles (UAVs) due to its high performance in complex tasks and relatively low online computational cost. Despite these advantages, the lack of interpretability in RL and decision-making processes poses significant challenges to safety verification, which severely limits the practical deployment of RL-based autonomous flight systems in safety-critical scenarios. Ensuring safety during both training and testing remains the key for real-world applications. To address this issue, a Run Time Assurance (RTA) filter framework for a UAV collision avoidance system based on RL is proposed. The framework continuously monitors the avoidance actions generated by RL and ensures that unsafe actions can be identified and corrected in real time. The framework provides security while retaining the performance advantages of RL. First, a discrete kinematics model of the UAV and a collision avoidance environment for a three-dimensional multi-intruder UAV are established. The Proximal Policy Optimization (PPO) algorithm is used to train the UAV avoidance policy. Secondly, the RTA filter with safe constraints is designed based on this model. To reduce the position error between the discrete kinematics model and the continuous dynamics model of the UAV, optimization is performed using the Sequential Least Squares Programming Method (SLSQP). The RTA is applied to the training and testing phases of UAV collision avoidance systems to monitor and correct unsafe actions online. Finally, three training and testing methods are set: no RTA, RTA without punishment, and RTA with punishment. Comparing different relaxation coefficients allows for analyzing their impact on filter performance. Simulation results demonstrate that the RTA effectively enhances safety throughout training and testing. Compared with the no RTA method, the collision rate is reduced by 7.04%, and the success rate is increased by 2.68%. In extreme collision avoidance scenarios, the average collision rate obtained by different reinforcement learning algorithms is reduced by 21.46%, and the average success rate is increased by 15.54%.
Ma, Zan, Yan, Liqin, Bai, Jie, Liu, Yubin, Sun, Shuguang
While Unmanned Aerial Vehicles (UAVs) hold immense potential for transforming the low-altitude economy, developing an agent that combines both intelligence and flexibility remains a significant challenge. To this end, we propose a novel aerial navigation system, Aeroseek, to search for executable paths that coherently transition from cognitive understanding to physical movement, enabling more autonomous and flexible navigation with instruction guidance. Specifically, given a navigation instruction, we follow a fast-andslow thinking procedure inspired by human pilots by breaking down complex decisions into manageable sub-tasks and initiating motion planning immediately upon the completion of each sub-task for realtime movement. By integrating a multi-branch Chain-of-Thought, Aeroseek is capable of combining reasoning with spatial recognition and self-verification to search for robust sub-goals under expected constraints. For sub-goal navigation, we introduce Online Kinodynamic Monte-Carlo Path Planning (OnKi-MCPP), a novel motion planning approach handling both kinodynamic constraints and path search, while ensuring obstacle avoidance for UAVs. Benefiting from the proposed two-tier searching protocol, Aeroseek demonstrates comprehensive capabilities in processing complicated instructions and adapts to dynamic aerial conditions. Extensive experiments demonstrate the advantages of Aeroseek across tasks of varying complexity.
Luo, Guoan, Xu, Tian, Zhou, Jiacong, Wang, Xianyun
In high-end aerospace mission scenarios such as distributed sensing and tactical reconnaissance, path planning for multi-UAV cooperative formation flight has become an indispensable core component of mission systems. However, a key unresolved technical bottleneck remains: how to achieve an optimal balance between strictly maintaining the UAV formation’s preset geometric configuration and realizing collision-free, flexible flight in dynamically changing complex environments. Multi-Agent Reinforcement Learning (MARL) can theoretically solve this coupled optimization problem, but the control instructions generated by the existing mainstream MARL methods often have high-frequency fluctuation issues. Direct deployment of these fluctuating commands on physical UAV hardware will increase alternating stress on mechanical structures, raise the risk of motor overheating, and significantly boost battery energy consumption. These problems directly restrict the engineering application of such methods. To address the above challenges, this paper proposes a robust control framework that accounts for UAV physical characteristics to solve the model mismatch between simulation environments and real flight scenarios during carrier-based UAV deployment. Firstly, to improve the learning efficiency of the algorithm and avoid the cold start problem in the initial training stage, this paper introduces the Adaptive Curriculum Learning (ACL) method. Unlike the fixed-difficulty training mode, ACL can dynamically adjust the environmental complexity based on the real-time task performance of the UAVs. It first allows agents to master basic formation maintenance skills, before training is gradually transitioned to more complex tasks such as navigation in intricate environments. Additionally, considering the computational limitations of the onboard embedded processors of UAVs, this paper only integrates the Graph Attention Network (GAT) into the centralized Critic network of the centralized training and decentralized execution (CTDE) framework. The core advantage of this design is that the system no longer solely relies on distance as the judgment basis but assesses the influence of adjacent UAVs based on the collision risk level. This not only improves the accuracy of value estimation, but also does not compromise the response speed of local control instructions- a feature that is crucial for the real-time requirements of UAV flight control. The proposed method was tested and validated in a high-fidelity simulation environment built in strict compliance with UAV kinematic constraints. The experimental results fully demonstrate its superiority: in dense obstacle areas, the success rate of the task reached 80.4%; compared with the aggressive baseline method, the jerk was reduced by 10.9%. Most importantly, the flight trajectories generated by this framework are smooth and adaptable to actual flight conditions, and the corresponding control commands can be directly deployed on real flight controllers without additional parameter adjustments. This result demonstrates that the proposed framework achieves substantial technical progress in enhancing the autonomous response capabilities of UAV clusters to complex scenarios.
Huang, Xingyu, Zhang, Jiyu, Jiang, Yanshu, Fan, Yuhang
This paper describes an intelligent unmanned aerial vehicle (UAV) system that automates red pine cone harvesting in dense forests to mitigate the safety and inefficiency of traditional manual harvesting. The suggested system combines multi-sensor fusion perception, variable-stiffness rigid-flexible composite manipulator, and smart path planning into a single aerial harvesting system. To create a three-dimensional model of the environment and robust cone detection under complicated lighting conditions and extreme canopy occlusion, a heterogeneous sensor suite such as LiDAR, depth cameras, and inertial sensors is used. According to the results of fused perception, collision-free approach paths are created to lead the UAV safely to target cones in limited three-dimensional spaces. One of the significant contributions of this work is the iterative design and optimization of a dedicated harvesting manipulator to be used in aerial forestry operations. A comparative analysis of various structural designs leads to the creation of a rigid flexible composite manipulator with an effective workspace diameter of 1.8 m and adequate stiffness to reduce the effects of the downwash of the UAV rotor. The disturbance-assisted harvesting strategy is adopted by the system, which enables efficient dislodgement of cones without requiring accurate grasping. The proposed technical route is proven to be feasible based on laboratory experiments and initial field tests carried out in real forest settings. It has been experimentally proven that the combined system can effectively perform autonomous harvesting cycles, including target recognition, navigation, and cone dislodgement, thus offering a practical technical paradigm of intelligent and safe forestry harvesting applications.
Shi, Shuaiqi, Du, Yisheng, Li, Xiang, Wang, Weidong
As the use of low-altitude drones in urban environments continues to grow, the likelihood of near-miss events and collisions between birds and drones operating in shared airspace has drawn increasing attention. To address this issue, this paper proposes a collision probability assessment framework based on the Boids swarm behavior model and Monte Carlo statistical methods, using the Chengdu silky-tailed starling as a case study: simulating the relative motion between bird flocks and UAVs within typical flight segments, defining near-misses and physical collisions events based on minimum separation distance thresholds, and employing Wilson confidence intervals to quantify uncertainty and analyze convergence. The simulation results indicate that after introducing the detectionavoidance mechanism, the minimum distance distribution function curve shifts rightward overall, collision duration is significantly reduced, the probability of event occurrence is effectively lowered, and the cumulative estimate exhibits good convergence under large samples. Additionally, the probability of incidents exhibits a monotonically decreasing trend with enhanced detection capabilities, revealing a monotonically sensitive characteristic where higher detection capabilities reduce collision probability risks. This provides a quantitative basis for low-altitude planning and risk management.
Pan, Anqi, Long, Yong, Li, Xiaojing, Chen, Di, Yang, Jibo, Zhao, Jinchen, Wu, Honggang
With the continuous advancement of collaborative unmanned aerial vehicle (UAV) swarm applications, the importance of intent recognition for aerial swarms in UAV management and control has become increasingly evident. To address the shortcomings of existing intent recognition methods, this paper proposes a knowledge-driven aerial swarm target intent recognition method designed for few-shot scenarios. First, a target trajectory clustering algorithm is designed by combining trajectory similarity measurement (which considers directional features) with density peaks clustering. This algorithm groups targets exhibiting the same or similar motion trends, thereby transforming the swarm target intent recognition problem into a swarm member intent recognition problem, which reduces the complexity of intent recognition. Second, considering the behavioral dynamics inherent in aerial swarm collaborative applications, a target intent inference model is constructed based on a Dynamic Bayesian Network (DBN). Utilizing the characteristic information of swarm targets at different time points, this model achieves intent inference and recognition through a hierarchical process: “sub-intent inference → accumulation of historical sub-intent sequences → intent inference.” Finally, simulation experiments are conducted to verify the feasibility and effectiveness of the proposed method.
Jiang, Jianxiong, Gui, Yang, Sun, Peng, Yang, Xinyu, Yang, Guangwei, Zhong, Yun
To solve the critical challenges of uncoordinated operational node selection and premature mission termination caused by dynamic energy or coating depletion during anti-corrosion spraying on largescale steel structures, this paper proposes an advanced path planning methodology for UAVs. The proposed approach is built upon a BIMSLAM fusion model. Specifically, a high-fidelity three-dimensional (3D) environmental model is reconstructed by integrating offline BIM structural priors with real-time SLAM data. Furthermore, the framework implements the ICP algorithm and Kalman filtering simultaneously to eliminate point cloud noise and compensate for perceptual occlusions, thereby achieving a 5 cm localization accuracy in cluttered industrial environments. A dual-stage pipeline is established for pre-screening and dynamic evaluation. In the offline phase, geometrically inaccessible task points are identified and pruned via spatial accessibility analysis. During real-time operations, the onboard system continuously calculates the battery State of Charge (SoC), operational power, and remaining paint volume to eliminate inefficient flights. For the path generation stage, a hybrid search strategy is developed, which implements a Genetic Algorithm for task sequencing alongside a multi-objective cost function parameterized by four metrics: energy consumption, paint requirements, obstacle proximity, and task density. Additionally, Softmax is used to dynamically regulate the weight coefficients, enabling the system to balance operational safety and efficiency based on the instantaneous state vector. Finally, a fifth-order B-spline parameterization is used to maintain flight stability under complex wind disturbances and heavy time-varying payloads. Consequently, the results show that the proposed method significantly outperforms conventional baselines. The recorded 96.7% mission completion rate demonstrates its operational reliability for the maintenance of bridges and towers.
Qian, Lei, Huang, Chen, Tan, Libin, Huang, Tao
This paper presents GBC-YOLOv5s, a lightweight vehicle detection model designed for unmanned aerial vehicle (UAV) applications in the growing low-altitude economy sector. By utilizing the baseline YOLOv5s architecture, we integrated three structural enhancements to achieve an optimal trade-off between algorithmic precision and processing expenses. First, ghost convolution is used to exploit feature map redundancy, substantially cutting both parameter count and computational overhead. Second, a Bi-Level routing attention (BRA) mechanism is integrated. This allows the model to capture global context along with fine-grained local features of vehicle targets effectively. Third, standard upsampling is replaced with content-aware reassembly of features (CARAFE). This better preserves feature details, which proves particularly important for detecting small objects. Testing was conducted on a custom UAV-based vehicle dataset. GBC-YOLOv5s achieves 86.6% average precision (AP) with only 4.7 M parameters. This represents a 34.7% reduction from the original YOLOv5s with 7.2 M parameters. The accuracy drop is merely 0.2%. Ablation experiments validate both the individual impact and synergistic nature of each proposed module. Results show that GBC-YOLOv5s reaches a practical balance between performance and efficiency. It offers a workable solution for real-time vehicle detection on resource-limited UAV platforms and intelligent traffic monitoring applications.
Zhang, Qingpei, Yu, Tianhao, Zhao, Zheming
Low-altitude networks, which support communication among ground stations and aerial nodes such as UAVs, require real-time monitoring to ensure reliable operations under dynamic conditions. The highly mobile nature of UAVs, coupled with frequent topology changes and limited onboard resources, presents unique challenges for network monitoring. In-band Network Telemetry (INT) provides a promising approach to monitor low-altitude networks by embedding telemetry metadata directly into data packets, enabling fine-grained visibility into network behavior. However, traditional INT schemes introduce substantial bandwidth overhead, a critical issue for resource-constrained low-altitude links. Furthermore, existing optimization approaches typically rely on static configurations, which cannot effectively handle the highly dynamic and volatile conditions we observe in low-altitude network environments. We present MGAT (Multi-Granularity Adaptive INT), a lightweight telemetry mechanism for low-altitude networks that achieves efficient monitoring through decoupling control decisions from data collection operations. Our data plane design exploits the spatiotemporal characteristics of different telemetry data types through three specialized collection schemes: flow-level aggregation is used for statistical metrics, static attribute sampling handles time-invariant properties, and packet-level extremum tracking captures critical outliers. The control plane combines QoS-aware flow classification with load-adaptive sampling, allowing the system to dynamically adjust monitoring intensity according to real-time network utilization. We evaluate MGAT through experiments and find that it reduces overhead by 98.7% compared to traditional INT while maintaining high detection accuracy (F1-score of 0.965). When network load varies dynamically, MGAT demonstrates 8.3% higher accuracy than static baseline methods and consumes 11.8% less bandwidth, confirming its ability to adapt effectively to the complex dynamics of low-altitude networks.
Wu, Zongyan, Hu, Jinju, Liu, Kang, Liu, Ying
In the field of unmanned aerial vehicle (UAV) signal recognition based on deep learning, complex electromagnetic environments pose severe challenges. Among them, interference suppression and recognition robustness under low signal-to-noise ratio (SNR) conditions are major research focuses. To tackle these issues, this paper proposes a robust radio frequency (RF) signal recognition method for UAVs. The method is built on an improved ResNet18 network. It separates the video transmission signals and control signals of UAVs, and adopts mixed-SNR training. These designs greatly strengthen the model’s ability to perform reliable recognition. First, the short-time Fourier transform (STFT) is used to convert UAV I/Q signals into time-frequency images. A refined time-frequency masking strategy is then adopted. This strategy separates and suppresses interference from control signals according to their bandwidth differences from video transmission signals, thus improving signal quality. Second, a squeeze-and-excitation (SE) channel attention module is embedded into each residual block of ResNet18. This structure enables the model to concentrate on critical features. The classification head is also rebuilt into a three-layer fully connected network to strengthen nonlinear transformation and complex feature interaction. Given the performance variation of the model across different SNR levels, a mixed-SNR training strategy is adopted. Fine grid search experiments are conducted to identify the optimal data distribution for maximizing recognition performance. Experimental results on the public DroneRFb-DIR dataset show that the improved ResNet18 model achieves a 3.07% increase in classification accuracy. The time-frequency masking method brings an extra performance gain of 1.08%. Although the mixed-SNR strategy causes a slight accuracy drop at the high SNR of 20 dB, it improves recognition accuracy by 8% to 53.01% within the low-SNR range of −10 dB to 10 dB. It also effectively reduces misclassification between different UAV types and enhances the model’s robustness. The proposed method provides a feasible solution for accurate and stable UAV signal recognition in complex electromagnetic scenarios.
Yu, Jiankun, Chen, Yongyou
The Unmanned Aerial Vehicle (UAV) swarm require efficient scheduling to maintain a high mission success rate in dynamic environments. At present, most methods are designed for static targets. But real buoys move and drift. The recommendation of “short-term drift theory” leads to the failure of multi-scheduling. We propose a Dynamic Hierarchical Group Control-Deep Reinforcement Learning (DHGC-DRL) framework based on hierarchical deep reinforcement learning. The framework follows a divide-and-conquer strategy and splits the problem into two parts. The first part is a global task assignment. The second is local path planning. Using the Coordinated Dynamic Task Allocation (CDTA) transformer in upper-layer building policies affect dynamic requests and quickly extracts information from the environment. This enables the model to find the best task assignment. Combinatorial Multi-Armed Bandit (CMAB) and the transformer module are used in the lower layer. These blocks modify the path and help the UAV adapt to the movement of the buoys. This makes the planning result more accurate. We also propose a coordination mechanism for the reward function. The mechanism uses a joint reward function, which allows two layers to exchange messages with each other. This can achieve real-time and fast solutions. The algorithm is tested by using the Double-Level Deep Reinforcement Learning (DL-DRL) baseline in parallel. The DHGC-DRL algorithm achieves faster convergence and improved training efficiency. The mission success rate increases by 25%. Path feasibility improves by 14.3% and overall performance is improved by 8.2% compared to the baseline.
Zhang, Xin, Du, Xuefeng
Low-altitude Communication Network (LCN) refers to network systems that employ drones and other low-altitude aerial platforms as carriers. Leveraging advanced communication technologies and data processing methodologies, these networks enable high-efficiency data transmission in low-altitude airspace, thus emerging as a pivotal driver for the advancement of the low-altitude economy. With the continuous advancement of Unmanned Aerial Vehicle (UAV), artificial intelligence, and communication technologies, LCN have been increasingly integrated into smart city construction, Internet of Things expansion, emergency management, and other fields. As a core component of these sectors, LCN have effectively driven the robust development of the low-altitude economy. Within the context of LCN, the efficient deployment of Service Function Chain (SFC) emerges as a critical factor in optimizing network service capabilities and enhancing resource utilization efficiency. In the dynamic and complex operating environment of LCN, reasonable and efficient SFC deployment can effectively coordinate the allocation of communication, computing, and storage resources, laying a solid foundation for the stable and high-quality operation of various network services. However, existing deployment algorithms either fail to fully consider the dynamic changes in the LCN or lack sufficient exploratory strategies, making them difficult to adapt to the diverse business scenarios within the LCN. To address the issue, our study proposes a dynamic SFC deployment method based on the Asynchronous Advantage Actor-Critic (A3C) and Graph Attention Network (GAT) algorithms. The experimental findings indicate that our proposed algorithm achieves a notable reduction in orchestration expenses, while simultaneously enhancing resource utilization, thus exhibiting excellent performance and robust adaptability in dynamic LCN environments.
Yang, Zihan, Yin, Jianhui, Yu, Chengxiao, Hu, Jinju, Liu, Ying
Flight positioning accuracy is a key technical indicator of unmanned aerial vehicle (UAV) systems. To quantitatively and accurately evaluate the positioning accuracy of UAV flight trajectories, this paper proposes a theoretical model for motion trajectory measurement and positioning, as well as a systematic solution for tracking measurement, based on a 2D tracking turntable carrying a dual-mode optoelectronic fusion integration system consisting of a laser ranging sensor and a vision camera. It analyzes the main influencing factors of measurement errors and establishes a measurement error model for flight trajectory positioning. Finally, flight trajectory tracking, positioning, and measurement experiments are carried out based on the principle prototype, which verifies the correctness of the method model and the feasibility of the system scheme. This study provides a theoretical basis and technical support for the testing of UAV flight control positioning accuracy indicators.
Qi, Weijie, Qiao, Yongming
The rapid growth of the low-altitude economy has accelerated the deployment of unmanned aerial vehicle (UAV) technology in a wide range of applications, including power infrastructure inspection and disaster relief operations. These scenarios impose stringent real-time processing demands on onboard vision systems. However, aerial images are characterized by a high density of small and distant targets, which significantly weakens feature representation quality, and current state-of-the-art detection models carry substantial computational overhead that conflicts with the limited hardware resources available on embedded UAV platforms. This paper addresses both challenges simultaneously by proposing a novel lightweight, end-to-end detection framework designated the High-frequency Enhancement Retentive Detection Transformer (HER-DETR). The framework introduces three coordinated technical contributions. First, a lightweight backbone network termed the Cross Stage Partial High-frequency Enhancement Network is constructed. This backbone integrates a cross-stage high-frequency residual enhancement mechanism to alleviate feature attenuation commonly observed in small-target scenarios. Second, a Depthwise Intrascale High-Low Feature Interaction block is designed for intra-scale feature interaction at the highest semantic level, strengthening fine-grained representation through a dual-frequency decoupling strategy. Third, a Retention Feature Fusion module is developed as the core cross-layer feature fusion component, enabling bidirectional multi-scale feature integration via a lightweight sparse attention mechanism. Comprehensive experiments on the VisDrone2019 and HIT-UAV benchmark datasets confirm that HER-DETR achieves state-of-the-art performance. Compared with the RT-DETR-r18 baseline, the proposed framework improves mean average precision at a 0.50 intersection-over-union threshold by 2.0 percentage points on VisDrone2019 and by 1.6 percentage points on HIT-UAV, while simultaneously reducing model parameter count and computational floating-point operations by 41.6% and 35.3%, respectively. These results validate the framework’s capability to balance detection accuracy and computational efficiency for real-time UAV deployment.
Jiao, Minghai, Liang, Yinghan, Zhang, Ende, Kan, Shuang
Currently, UAVs are increasingly used in remote sensing monitoring, logistics and distribution, emergency rescue and recreational aerial photography, etc., which puts higher requirements on wireless networks, i.e., how to allocate limited wireless resources in real-time, efficiently and fairly to ensure that a large number of UAVs can work together in complex environments and maintain high-quality communication services. This paper investigates an optimisation method for UAV scheduling based on the BeiDou satellite navigation system, which improves the accuracy and efficiency of UAV mission execution by utilising the high-precision positioning and time synchronisation services provided by the BeiDou system. The results of the experiment showed that as the number of missions increased from 10 to 50, the average mission completion time increased from 8.3 to 12.5 minutes, the average speed increased from 30 to 34 km/h, the average energy consumption increased from 0.8 to 1.2 kWh, the average distance flown increased from 12 to 22 km, the communication success rate decreased slightly from 99.2 to 98.2%, the resource utilization rate increased from 85.1% to 87.5%. These results show that the method performs well in terms of task completion time, energy consumption, communication success rate and resource utilisation, which provides strong support for the application of UAVs in the fields of logistics and distribution, emergency rescue, and so on.
Ji, Hongbin, Jiang, Kexin, Zhang, Dongxue, Bai, Yanhui, Liu, He, Zhang, Shuangna
In order to support successful tracking and monitoring tasks in different environments, drones need to quickly and accurately identify targets. This paper studies a drone combat target positioning method, with a focus on the BeiDou satellite navigation system and its performance under different environmental conditions. A complete system has been designed, including flight control system, environmental sensors, data transmission and processing system, aiming to achieve precise positioning and target tracking of unmanned aerial vehicles. This paper uses on-site testing to explain the coverage range and stability positioning accuracy of ranging signals in different environments. Experimental data shows that the BDS system has good positioning error performance for unmanned aerial vehicles, with an average error of 46.52 cm. BDS positioning time is usually short, with a maximum positioning time of 780ms and a minimum positioning time of 224ms. However, the BDS system still needs improvement in terms of dynamic environmental adaptability, multi-target tracking, anti-interference ability, and adaptability to harsh environmental conditions. Future work will focus on algorithm optimization, hardware technology improvement, and data security to enhance system performance.
Tao, Zhiping, Zhang, Chenghao, Hou, Bo, Song, Juan, Yuan, Peng, Liu, Xianyang
The present study concentrates on aerodynamic behavior and static stability characteristics, and propeller slipstream interference of twin- boom unmanned aerial vehicles (UAVs) with three typical tail configurations. This study adopts a coupled VSPAERO approach, which incorporates a VLM-based solving technique and the Actuator Disk Model. Simulation results under the no-slipstream baseline condition reveal that the inverted V-tail layout delivers the highest aerodynamic efficiency. Its lift/drag ratio is considerably superior to those of the U-tail and inverted U-tail designs. While all configurations maintain good stability in longitudinal and directional axes, only the U- tail exhibits lateral static stability. When propeller slipstream is introduced, the lift-to-drag ratio of all layouts decreases noticeably, mainly due to strong local low-pressure regions that substantially increase pressure drag. In addition, the slipstream has little influence on longitudinal static stability, leaving pitching moment characteristics nearly unchanged. Nevertheless, the swirling flow generated by the propeller effectively improves lateral stability. It enhances the stabilizing effect of the U-tail, turns the lateral instability of the inverted U-tail into stable behavior, and reduces the lateral unstable characteristics of the inverted V-tail. In addition, the findings show that the sideslip angle β’s impact on the yawing moment is affected by the angle of attack α. Coupling between α and β, as well as propeller slipstream, should be considered in directional stability analysis. Increasing propeller diameter appropriately improves aerodynamic efficiency and directional static stability, while placing the propeller at or slightly forward of the baseline position minimizes adverse tail interference.
Wen, Jiayi
To address geometric precision degradation and tracking divergence in single TDOA or AOA systems, this paper proposes a multi-station UAV tracking method integrating TDOA/AOA hybrid localization with nonlinear filtering to overcome precision degradation and tracking divergence in single-sensor systems. A Weighted Least Squares (WLS) method via pseudo-linearization provides reliable initial coordinate estimations, followed by an Unscented Kalman Filter (UKF) for continuous trajectory tracking. Simulations demonstrate that the proposed architecture eliminates network blind zones and robustly handles abrupt target maneuvers. Compared to the standard EKF, it rapidly suppresses peak RMSE to ~22 m and achieves a steady-state error of 10-15 m. This approach offers a high-precision, rapidly converging passive tracking solution for urban low-altitude defense.
Shen, Huaming, Wang, Yiting, Lv, Zhenbin, Zhou, Haosu, Lu, Wenbin
Conventional unmanned aerial vehicle (UAV) landing gears exhibit deficiencies including severe stress concentration, insufficient cushioning deformation capacity, and inadequate fatigue resistance during the landing impact process, which makes it difficult for them to meet the technical requirements of high-overload buffering conditions. Based on the elastic mechanics theory of the cantilever beam and using a two-degree-of-freedom (2-DOF) flexible hinge as the base, an analytical equation for the rotational stiffness of a 2-DOF flexible hinge in the direction was established. By combining the response surface methodology (RSM), the influence laws of key geometric parameters on the mechanical performance of the hinge were systematically analyzed. The optimal value range of the core structural parameters of this 2-DOF flexible hinge was determined as follows: the radius of the rounded corner R = 2.5 - 3.5 mm, the minimum thickness t = 1.5 - 2 mm, and the width a = 20 - 30 mm. Building on this foundation, an axial gradient columnar buffering structure incorporated with flexible hinges is proposed for the application of UAV landing buffering. Cross-verification between theoretical analysis and finite element simulation shows that the relative errors of structural deformation in each loading direction are merely 1.76% and 1.9%, respectively, which validates the accuracy and reliability of the proposed theoretical model. Parametric studies incorporating transient dynamic analyses under impact loading conditions elucidate that the maximal deflection and peak stress characteristics of the flexible hinge are positively correlated with the drop height and the structural connection angle. The outcomes of this research contribute a foundational theoretical framework and actionable technical support for facilitating the lightweight design and performance enhancement of flexible landing gears within high-load buffering systems of UAV.
Zhang, Qixian, Zhou, Jinhua, Zhang, Dongsheng, Wang, Jiashuai, Yang, Zengliang, Zhang, Yingxue
Time Difference of Arrival (TDOA) combined with the Extended Kalman Filter (EKF) is a widely used approach for dynamic target localization. However, in practical UAV localization scenarios, the EKF may suffer from divergence caused by nonlinear measurement characteristics, initialization errors, and measurement disturbances. To improve the robustness of EKF-based TDOA localization, this paper proposes a threshold-constrained EKF method, in which the velocity components of the updated state are limited within a physically reasonable range. The proposed strategy suppresses abnormal state updates and improves tracking stability under noisy conditions. Simulation results under straight, square, and circular trajectories, as well as outdoor experiments with real UAV signal data, show that the proposed method achieves better localization accuracy than the conventional TDOA-EKF. In the real-world tests, the mean localization error is reduced by 38.31% and 74.36% for two flight paths, respectively. The proposed method provides a simple and practical way to improve the robustness of TDOA-based UAV localization.
Lv, Zhenbin, Li, Yingying, Shen, Huaming, Zhou, Haosu, Wang, Anping, Zou, Weiwen
An adaptive equivalent consumption minimization strategy (A-ECMS) is proposed for a hydrogen fuel-cell range-extended UAV hybrid power system composed of a PEM fuel cell, a lithium-ion battery and a supercapacitor. The equivalent factor is updated online by a PID-based SOC feedback mechanism to improve the adaptability of conventional ECMS under variable flight loads. Under the same mission profile and the same initial and terminal SOC constraints, A-ECMS consumes 4.6277 g of hydrogen, reducing hydrogen consumption by 46.90%, 67.51% and 58.42% compared with PID control, conventional ECMS and state-machine control, respectively. Fuel-cell power fluctuation and battery SOC deviation are also reduced.
Li, Mengda, Liu, Zitong
The low-altitude economy has been established as a strategic emerging industry, driving the development of various new aircraft types, including electric vertical take-off and landing (eVTOL) vehicles and industrial unmanned aerial vehicles (UAVs). In this development, proton exchange membrane fuel cell (PEMFC)/lithium-battery hybrid propulsion systems face challenges in balancing fuel economy, stack durability, and implementability under stochastic aerodynamic loads. Six energy management strategies (EMS) are systematically evaluated in this study, namely Thermostat, Fuzzy Logic, State Machine (SM-EMS), Power Following (PF-EMS), Dynamic Programming (DP-EMS), and a proposed load-disturbance-based method (LD-EMS). These strategies are tested on three distinct platforms, including a 2 kW fully simulated system, a 2 kW hardware-based setup incorporating a real PEMFC stack and Li-FePO4 battery, and a 30 kW hybrid testbed. Experiments were conducted under synthetic stochastic load profiles representative of gust-induced power transients. The results indicate that LD-EMS achieves a total equivalent hydrogen consumption of 2.71 g (close to the DP-EMS optimum of 2.64 g and 19 % lower than SM-EMS) while reducing fuel cell state-of-health degradation to only 0.0005 % over the test cycle, which is an order of magnitude better than PF-EMS and SM-EMS. On the 30 kW platform under two distinct challenging profiles, LD-EMS maintains fuel cell efficiency above 90 %, keeps the stack operating in high-efficiency zone, and limits power switching events to 97 and 212 respectively over test cycles, demonstrating robust scalability and transient suppression capability without requiring prior mission knowledge. A five-dimensional evaluation framework is introduced to compare the six EMS. In accordance with this framework, platform-specific selection guidelines are established. Thermostat EMS is recommended for small industrial UAVs where hardware simplicity dominates; LD-EMS is the optimal choice for passenger eVTOL under unpredictable gust conditions; and SM-EMS is preferred for commuter/regional aircraft operating on fixed, repetitive routes.
Gao, Zibo, Chigan, Dongdong, Xiao, Yizhuo, Li, Yuansheng, Ji, Guoming, Han, Qing, Li, Zhi-peng
This paper studies how scale-free correlations, a key feature for efficient group response, emerge in collective motion under communication energy constraints. A two-dimensional Boids model integrated with a point-to-point wireless communication energy consumption model is used. The model dynamically determines, based on a limited communication power budget, an individual’s interactive neighbors, where neighbors are selected based on two different mechanisms: distance-based and probability-based, until the total communication cost reaches the budget. Simulations of flock sizes ranging from 30 to 230 individuals confirm that scale-free correlation persists, even under strict power constraints, which is shown by a correlation length ξ proportional to the flock’s linear size L. Notably, the ratio ξ/L, which reflects the robustness of collective correlation, displays a non-monotonic dependence on the available communication power. Initially, increasing power enhances ξ/L by expanding effective interaction ranges and improving information transmission. However, beyond a threshold, further power increase causes a slow drop in ξ/L, likely due to signal interference from excessively many or distant connections. This finding demonstrates that communication power is a critical, tunable engineering parameter regulating collective coordination, analogous to interaction range or neighbor number in biological models. The results imply that for engineered swarms, maximizing communication resources is not always beneficial; instead, an optimal power allocation exists to maximize coordination performance, balancing connectivity with efficient, localized interactions.
Zhao, Shiping, Gan, Qingzhong, Liu, Jinglei, Liu, Zhihao, Zhang, Liang
With the unprecedented surge in demand for flexible deployment and ultra-high throughput in next-generation wireless networks, airborne base stations (ABS), particularly unmanned aerial vehicles (UAVs), have become emerged as indispensable mobile communication infrastructure. To facilitate high-gain directional transmission within the millimeter-wave band, ABS frequently employ massive MIMO antenna arrays. However, under random disturbances caused by non-stationary turbulence, the system faces a severe physical bottleneck: weak anti-interference capability due to narrow beamwidth, which can easily lead to a sharp drop in link capacity. Addressing the two major shortcomings of existing deep learning prediction schemes—susceptibility to sensor noise interference and negative capacity performance shifts during stable periods, and response lag due to memory inertia during sudden turbulence periods—this paper proposes an ARET-ALSTM active compensation network based on SNR-aware multimodal attention, agile reset and adaptive event-triggered mechanisms. This network incorporates an event-triggered mechanism governed by a dynamic physical variance threshold, adaptively executing compensation during stable intervals to prevent excessive overshoot and redundant computational overhead. Simultaneously, it employs an agile feature reset mechanism based on an adaptive sliding window for concept drift detection, forcibly discarding old memories and reconstructing attention weights during sudden wind shear, achieving millisecond-level ultra-fast adaptation and directly driving network optimization with the goal of maximizing communication capacity. Simulation experiments show that, under different array configurations ranging from 2×2 to 16×16 and complex flow conditions, ARET-ALSTM successfully overcomes the physical capacity collapse limitation under large antenna scales, and its tracking agility and robustness are significantly better than benchmark algorithms such as Transformer, RNN and traditional Kalman filtering.
Tu, Jiajun, Zhou, Jihua, Li, Yanhang, Zhao, Tao, Huang, Hua, Wen, Yuanbin, Guo, Ziwen
Due to the limitation of unstable wireless links and resource-constrained edge computing device, it is challenging to transmit the unmanned aerial vehicles (UAVs) captured images with large data volume in real-time. To address this issue, this paper proposes a lightweight hybrid compression method named Jpeg Plus Plus Network (JppNet) for efficient multimodal image transmission. Different from existing methods that mainly focus on RGB image compression for offline storage and often use large networks for coding and reconstruction, the proposed JppNet focuses more on edge deployment under limited computing resources. At the encoder side, JPEG is first used to perform high-ratio compression on the RGB, D, and IR channels. Then, a lightweight convolutional encoder network extracts supplementary binary features for progressive transmission. At the decoder side, a medium-scale reconstruction network fuses the supplementary code with the JPEG reconstruction to restore image details. To verify the effectiveness of the proposed method, this paper built a multimodal data acquisition prototype and collected 20,000 high-quality RGBD-IR image samples for training and testing. Experimental results show that JppNet combines the high coding efficiency of JPEG with the detail recovery ability of convolutional neural networks. With low computational cost, JppNet obtains much lower reconstruction errors than JPEG on the RGB, D, and IR channels, especially at very low compression ratios. This supports the efficient transmission of UAV captured multimodal images under limited communication conditions.
Pi, Haoran, Ao, Yile, Li, Zhi, Tang, Gang, Jiang, Bowu
The precise detection and localization of 'low-slow-small' (LSS) unmanned aerial vehicles (UAVs) pose significant challenges in terminal defense scenarios. These non-cooperative targets have extremely sparse representations in radar point clouds because of their minimal radar cross-sections and appear as textureless, blurred patches in visual images. Furthermore, traditional multi-modal fusion paradigms rely heavily on rigid geometric projection matrices, which frequently fail under the mechanical vibrations, thermal deformations, and extrinsic perturbations inherent in rapid deployment environments, which often results in severe feature misalignment and contamination. To overcome these critical bottlenecks, we present an end-to-end cross-modal detection framework built on Bidirectional Collaborative Energy Fields (Co-Field). By redefining the 3D object detection task as an energy field reconstruction process within a latent multimodal feature space, we entirely remove the need for absolutely precise extrinsic parameters. Our framework uses a novel bidirectional mutual-guided interaction mechanism. Specifically, a vision-guided radar branch employs deformable cross-attention to adaptively aggregate highly discrete radar geometric features around visual semantic centers, achieving a feature-level soft alignment that implicitly compensates for calibration drift. Concurrently, a radar-corrected vision branch utilizes radar-derived geometric occupancy probabilities as depth anchors to gate and calibrate visual features, effectively suppressing the inherent divergence of monocular depth estimation. Ultimately, the framework regresses the target's precise 3D states and fine-grained categories directly from the extrema of the reconstructed collaborative energy field, circumventing complex heuristic post-processing. Experiments on the curated MMAUD dataset show that the proposed Co-Field method achieves better robustness. It remains highly effective even under severe calibration errors up to 10 degrees and achieves a 74.2% recall rate at extreme sparse distances ranging from 70 to 100 meters, and consistently outperforms state-of-the-art rigid and soft fusion baselines.
Peng, Chao, Shao, Peng, Ding, Qiao, Chen, Xiaoqing, Han, Aojia, Xu, Yuelei
To improve the accuracy and safety margin deficiencies of unmanned aerial vehicles in autonomous medical emergency search-and-rescue missions, this paper develops an adaptive embedded fusion optimizer named CTCM-ALA for rescue UAV path planning. The proposed algorithm hybridizes the tribe-competition member-cooperation strategy and artificial lemming algorithm to construct hierarchically adaptive initial populations, speed up convergence and prevent local convergence traps. Tests on 12 CEC2017 functions prove its better optimization performance. Real-world applications in dense urban and complicated topographic medical rescue cases verify its reliability and precision.
Chen, Yuhan, Xiong, Junhui, Li, Yuesheng, Zhou, Yifan, Yao, Junjie, Li, Rujun
A switched adaptive flight control architecture is proposed for fixed-wing carrier-sub UAVs to handle abrupt mass and inertia variations, as well as parameter uncertainties induced by transient aerodynamic interference, during sequential deployment. The separation process is rigorously modeled as a switched linear system, and a combined direct and indirect adaptive control scheme featuring an online Recursive Least Squares (RLS) modification is designed to compensate for model uncertainties. Furthermore, a dwell-time admissible switching law is established by solving Linear Matrix Inequalities (LMIs), ensuring the stability of the switching process through theoretically guaranteed Lyapunov energy drops at switching instants. Numerical simulations verify that transient tracking error spikes are significantly suppressed, and high-frequency oscillations are effectively attenuated compared to baseline controllers.
Chi, Haiyi, Yuan, Shuai
To effectively address the challenges of cooperative control for quadrotor formations in complex environments, this paper extends the dual-quaternion based nonlinear model predictive control (DQ-NMPC) framework to leader-follower formation. Unlike conventional methods that treat translation and rotation separately, the dual-quaternion based model achieves a unified and compact representation of rigid-body motion using only eight real numbers without singularities, enabling global description on SE(3). Through the logarithmic map, the pose error is mapped from the Lie group to its Lie algebra, resulting in a quadratic cost function that naturally resolves the dimensional inconsistency between position and attitude errors, thereby simplifying weight tuning and improving numerical stability. Building on this formulation, a UAV formation controller is developed, where the relative motion between leader and follower is concisely described using dual quaternions, and a minimum safe distance constraint is explicitly incorporated to avoid collisions and maintain safe separation. In formation scenarios, the follower autonomously accomplishes chasing, rendezvous, and formation keeping under various trajectories, including straight-line pursuit and Lissajous curves. Moreover, the stability of formation control is validated under non-ideal conditions such as communication delay, mass mismatches, and constant external disturbances. Furthermore, the optimal weight combination is determined through grid search, which further improves formation tracking accuracy and control smoothness. In addition, the proposed framework inherently supports time-varying formation patterns, such as dynamic switching between different relative configurations and adaptive scaling. Computational efficiency analysis shows that the average solution time per control cycle is far below the sampling interval, validating real-time feasibility. Overall, the DQ-NMPC framework provides a unified, robust, and efficient solution for quadrotor formation control under both ideal and perturbed environments.
Li, Jiaye, Zhang, Zexu, Liu, Jingzong
Large-scale UAV clusters make real-time reconstruction of joint spatio-temporal-frequency (STF) situations difficult because the monitored state is high-dimensional and strongly coupled. This paper proposes a lightweight reconstruction algorithm based on Kronecker Compressive Sensing (KCS). First, a Multi-Channel Random Demodulation (MCRD) sensing model is built, and a Spatio-Temporal-Frequency 5D Grid (STF-5D Grid) is used to represent space, frequency, and time resources in a unified form. Next, a Tensor Orthogonal Matching Pursuit (T-OMP) algorithm is developed under the KCS framework. By using the separability of Kronecker products, the method replaces storage and multiplication of a global high-dimensional sensing matrix with dimension-wise operations. Two simulation studies are conducted to evaluate computational feasibility and reconstruction quality. In the city-scale case with 62.91 million grid cells, the factorized implementation avoids constructing the global sensing matrix and completes the tested reconstruction in 9.44 s. The reported memory value is interpreted as additional algorithmic storage for the factorized operators rather than total process peak memory, and should be rechecked with full tensor input/output and runtime overhead included. In noise-free, highly sparse proxy tests, several reconstruction errors approach the numerical precision floor; these results are therefore treated as ideal upper-bound evidence rather than practical field accuracy. The simulations suggest that KCS-based T-OMP can reduce storage and computational pressure for STF reconstruction, while its engineering performance still requires validation under noisy observations and realistic platform constraints.
Tian, Jinghao, Wang, Qianzhe, Di, Shen, Wei, Tan, Song, Bowen
Space-Air-Ground Integrated Network (SAGIN) emerges as an appealing framework for implementing pervasive connections among diverse layers of space, air, and ground. The scheme relies on satellite deployment for extensive space coverage, Unmanned Aerial Vehicle (UAVs) acting as agile aerial relay stations, and ground robots for performing the final task assignment. Nevertheless, integrating heterogeneous nodes with varying mobility characteristics poses significant challenges: satellite operations are restricted by orbital dynamics, ground robot movements are restricted by topography constraints, whereas UAVs demonstrate flexibility and can establish line-of-sight connections with both satellites and ground robots. Current solutions usually adopt UAVs merely as passive relay nodes, causing inefficient adaptation mechanisms and excessive communication overhead when responding to dynamic perturbations or blocked connections. To overcome the aforementioned challenge, this paper proposes a novel fuzzy initiative-yielding coordination approach assigning two opposite behaviors. Initiative is assigned to UAVs in order to facilitate coordinated operations between layers, whereas Yielding refers to the adaptive response of satellites and ground robots. Contrary to previous research where behavioral assignments follow strict threshold rules, our fuzzy inference considers various parameters such as link quality, node state, and task priority, providing flexible behavioral assignments. Our predictive approach incorporates the Exponentially Weighted Moving Average (EWMA) technique to perform proactive failure avoidance. Based on five comprehensive experiments with 30 independent runs each, our approach obtains similar task completion rates (84.4%), but exhibits significantly lower communication overhead by up to 11 to 147 times compared to baseline methods. Furthermore, the proposed behavioral assignments avoid any coordination conflict (0 vs. 30.8), reduce completion time by 10.1%, maintain robust performance under 20% rule reduction, scale to 150-node mega-constellation scenarios, and exhibit higher fault tolerance under dynamic environments (81.5% vs. 79.8% of CBBA under high stress conditions).
Bao, Jie, Zhang, Yuping, Zhang, Ronghao
Items per page:
1 – 50 of 1628