Fixed-time bipartite flocking of perturbed networked UAV systems: A distributed optimization approach

Integration paradigm of intelligent digital twin into UAVs systems

Breaking the low-cost barrier: a memory-augmented reactive navigation system for UAVs in cluttered indoor environments

AI-Driven UAV Systems in Precision Farming: Technologies, Applications, and Future Perspectives

An Edge-AI enabled UAV system for site-specific application targeting Palmer amaranth in corn and soybean fields

Abstract Fog-related flight disruption is costing big international airports more than Rs 2.5 crores for each such event, while the traditional countermeasures, chemical seeding and thermal heating, are expensive, slow and environmentally damaging. This paper proposes the first field-validated airport fog dispersal autonomous UAV system that combines deep reinforcement learning with targeted UV-C photolysis technology. Conventional ways of fog dispersal take 30–45 min for runway clearance, cost Rs 15,000 per operation and produce 500 kg of CO2 emissions. These strategies evaporate fog droplets without tackling the condensation nuclei that are causing them and so the fog can quickly reform. We use a 4-UAV swarm with UV-C LED arrays (254 nm wavelength) for the degradation of hygroscopic aerosols which act as cloud condensation nuclei (CCN). Unlike thermal approaches that only evaporate droplets, our photolysis-based approach can reduce the efficiency of CCN by 35–45% so that the fog does not re-form. A deep Q-Network (DQN), based on 625-256-256-8 architecture, autonomously coordinates swarm positioning based on real-time sensor fusion from LiDAR (25 × 25 m resolution), thermal imaging (640 × 480 at 30fps) and meteorological arrays. 96% accuracy of fog detection. Our operational flights took place at Sri Guru Ram Dass Jee International Airport Amritsar, India, with 120 flights starting from November 2024 till March 2025. Results show: 83.1% reduction in time of fog clearance (from 30 to 5.06 min), 80% improvement of runway visibility range (from 450 to 810 m), 95% reduction in cost (from Rs 800 to Rs 15000 per sortie), 96% reduction in CO2 emission (from 20 to 500 kg per operation).Randomized complete block design using Friedman analysis (kh2 = 128.45, p < 0.001, Cohen’s d effect sizes of 4.85–7.92 show very large practical significance for all of the metrics. Zero incidents during 120 flights with ground exposure from UV-C (0.008 mJ/cm 2 ) 375x below ICNIRP occupational limits. Real-time DQN inference latency (183+-27ms) is the aviation safety-critical requirement (< 250ms). This research sets up a scalable paradigm for fog management at fog-prone airports anywhere in the world economically and environmentally and the potential savings is Rs. 2.5 Crores every year at major international airports.

Semantic Communication (SC), driven by a deep learning (DL)-based "understand-before-transmit" paradigm, transmits lightweight semantic information (SI) instead of raw data. This approach significantly reduces data volume and communication overhead while maintaining performance, making it particularly suitable for UAV communications where the platform is constrained by size, weight, and power (SWAP) limitations. To alleviate the computational burden of semantic extraction (SE) on the UAV, this paper introduces federated learning (FL) as a distributed training framework. By establishing a collaborative architecture with edge users, computationally intensive tasks are offloaded to the edge devices, while the UAV serves as a central coordinator. We first demonstrate the feasibility of integrating FL into SC systems and then propose a novel solution based on Proximal Policy Optimization (PPO) to address the critical challenge of ensuring service fairness in UAV-assisted semantic communications. Specifically, we formulate a joint optimization problem that simultaneously designs the UAV's flight trajectory and bandwidth allocation strategy. Experimental results validate that our FL-based training framework significantly reduces computational resource consumption, while the PPO-based algorithm approach effectively minimizes both energy consumption and task completion time while ensuring equitable quality-of-service (QoS) across all edge users.

Abstract The shift towards renewable energy sources necessitates a stable supply of essential raw materials and calls for advanced, non-invasive techniques to explore deep-seated mineral resources. In the last years, the semi-airborne electromagnetic (sAEM) method has been further developed to efficiently investigate the subsurface down to ~1 km depth. The Kropfmühl graphite deposit in Lower Bavaria comprises high-grade metamorphic graphite-bearing gneisses. While mining reaches depths of 200 m, the western extent of the deposit remains unknown. To verify the western extension and continuity of the Kropfmühl graphite deposit, a drone-based semi-airborne electromagnetic survey was conducted, combining a dense spatial sampling of magnetic fields using an UAV system and a high signal strength with controlled-source electromagnetic (CSEM) transmitters on the ground. The survey area spanned approximately 1.8 × 2.5 km2, using four galvanically coupled dipole transmitters and multiple UAV flights over four days. To increase depth sensitivity, a total-field magnetometer alongside a three-component induction coil system was utilized. To derive a reliable subsurface conductivity model, a 3D inversion of multi-frequency, multi-source sAEM data of both vector and scalar magnetic field data was conducted. The obtained model revealed three east-west trending conductive anomalies, interpreted as graphite-rich zones. The main anomaly (C1), ~200 m south of the active mine, indicates possible fault-controlled displacement. Borehole resistivity logs and available HEM data correlate well with conductive zones and graphite enrichment. The results highlight the potential of UAV-based sAEM to image complex subsurface structures. The method delivers high-resolution, cost-effective data acquisition, with a high data coverage and only the transmitters requiring ground installation. This renders the method suitable for imaging geologically complex settings in hard-to-access areas.

A possible classification for organization purposes: [...]

This paper intends to address the challenges of insufficient robustness and model uncertainty compensation in unmanned aerial vehicle dynamic systems under complex disturbances. The paper proposes a hybrid control architecture that combines deep fusion model predictive control with adaptive Proportional–Integral–Derivative (PID) based on Transformer attention mechanism. The core innovation of this architecture lies in introducing attention neural networks to dynamically tune PID gains online, and forming a deep collaborative control framework of "prediction-learning-compensation" with Model Predictive Control (MPC) and sliding mode disturbance observer with H∞ (H-infinity) robust optimization. This thereby improvs the adaptability and control accuracy of the system under unstructured disturbances and model mismatches. The control architecture employs a robustly optimized upper-layer MPC controller, which, based on the receding horizon principle, utilizes real-time system state updates to predict future state evolution. An H∞ performance criterion is incorporated into the control sequence optimization to strengthen robustness against model parameter perturbations and external disturbances. The lower-layer controller adopts an adaptive PID structure that responds quickly to the reference signals generated by the MPC. To address the degradation of PID tuning performance under dynamic mismatches and unmodeled disturbances, an attention mechanism neural network based on the Transformer architecture is introduced to adjust the PID gains online and capture nonlinear dynamic variations. Additionally, in order to further enhance system stability under severe disturbances, this control framework integrates sliding mode control technology into the disturbance observer design, and constructs a sliding mode disturbance observer module for explicit estimation of external disturbances and model uncertainties. The estimated values are injected into the lower-level adaptive PID controller through a feedforward compensation mechanism to achieve active disturbance rejection. Simulation experiments conducted in a nonlinear disturbance environment built on the AirSim platform, as well as tests using the EuRoc dataset, demonstrate that the proposed method maintains a steady-state tracking error within 5% during path-following tasks. Compared with the traditional MPC combined with fixed gain PID control, this method improves the steady-state robustness by about 17%, and shortens the system adjustment time from 3.15 s to 2.47 s, significantly improving by 21.6%, demonstrating excellent convergence and anti-interference ability. The results indicate that the MPC-PID hybrid control approach offers significant advantages in enhancing the robustness, adaptability, and control accuracy of UAV systems, making it well-suited for intelligent control demands in complex flight missions.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-32436-6.

DRL-based power optimization for hybrid FSO/THz-enabled UAV systems using IR-HARQ

Design and simulation of a fuel cell–battery hybrid power system for extended-endurance quadrotor UAVs

Modeling and Unilateral Adaptive Control of a Flexible Slung Load System for Multirotor UAV With Actuator Constraints and Faults

Twistor-based integrated double hyperbolic sliding mode control for time-varying quadrotor UAV systems

Integrated Task Scheduling, Path Planning, and Control for Cooperative UGV–UAV Systems via an Extended MPPI-GA Framework

Autonomous UAV systems with few-shot visual intelligence for process-aware construction monitoring

Abstract. Accurate estimation of forest structural and taxonomic parameters is vital for biodiversity monitoring, carbon accounting and sustainable management. Most of the current methods for estimating these parameters are still developed and tested on site-specific case studies, limiting reproducibility and cross-site generalization. This paper introduces 3D3, a multi-sensor and multi-resolution benchmark dataset designed to evaluate 3D forestry algorithms across diverse European forest types. 3D3 includes data collected by airborne, helicopter, UAV and terrestrial (static and mobile) laser scanning systems along with RGB and hyperspectral imagery, covering a variety of forest types (Boreal, Alpine and Mediterranean). By encompassing both mono- and multi-wavelength laser data, 3D3 represents a unique resource for developing new algorithms and evaluating them on distinct datasets. Each site provides ground truth for at least one task among Individual Tree Segmentation (ITS), Forest Semantic Segmentation (FSS) or tree parameter estimation and species classification.

This paper proposes a deep learning framework for UAV-assisted cellular relay communication to optimize throughput, latency, and energy efficiency in dynamic urban environments. The framework integrates a Convolutional-Gated Hybrid Network (CGHN) for channel representation and a two-stage reinforcement learning (RL) algorithm. CGHN eliminates explicit Channel State Information (CSI) dependency by mapping raw pilot symbols into compact semantic vectors, reducing computational overhead. The RL approach combines a global Soft Actor-Critic (SAC) agent for coarse trajectory/power optimization and a local Proximal Policy Optimization (PPO) agent for fine-grained beamforming. Validated via 3GPP-compliant simulations and field tests, the method achieves a 35% throughput gain and 20% energy efficiency improvement over static UAV relaying, with low inference latency (2.7 ms) and scalable performance. Results demonstrate the potential of deep learning and hierarchical RL in enabling adaptive UAV systems for 6G air-ground networks.

A DQN-based algorithm is presented in this paper for maximizing throughput and providing quality of service (QoS) in an RIS-assisted UAV system. The system includes a UAV-BS coordinating multiple mobile users in a dynamic 3D environment. By utilizing DQN to dynamically optimize the trajectory, transmission power, and user scheduling online, the system is able to accommodate the changes of the environment and to achieve the maximum performance while keeping QoS. The proposed system model integrates an RIS element to optimize signal quality and coverage, especially in NLoS cases, by additionally enabling diversity. The performance of the proposed method is coordinated with the throughput maximization, the power efficiency , and the QoS guarantee in extensive simulations against traditional optimization techniques. It is shown that the proposed scheme has better adaptability and efficiency compared to traditional schemes in a complex and dynamic communication environment.

This study develops an automatic dispatch system for quadrotor UAVs that integrates air-writing gesture recognition with a graphical user interface (GUI). The DJI RoboMaster quadrotor UAV (DJI, Shenzhen, China) was employed as the experimental platform, combined with an ESP32 microcontroller (Espressif Systems, Shanghai, China) and the RoboMaster SDK (version 3.0). On the Python (version 3.12.7) platform, a GUI was implemented using Tkinter (version 8.6), allowing users to input addresses or landmarks, which were then automatically converted into geographic coordinates and imported into Google Maps for route planning. The generated flight commands were transmitted to the UAV via a UDP socket, enabling remote autonomous flight. For gesture recognition, a Raspberry Pi integrated with the MediaPipe Hands module was used to capture 16 types of air-written flight commands in real time through a camera. The training samples were categorized into one-dimensional coordinates and two-dimensional images. In the one-dimensional case, X/Y axis coordinates were concatenated after data augmentation, interpolation, and normalization. In the two-dimensional case, three types of images were generated, namely font trajectory plots (T-plots), coordinate-axis plots (XY-plots), and composite plots combining the two (XYT-plots). To evaluate classification performance, several machine learning and deep learning architectures were employed, including a multi-layer perceptron (MLP), support vector machine (SVM), one-dimensional convolutional neural network (1D-CNN), and two-dimensional convolutional neural network (2D-CNN). The results demonstrated effective recognition accuracy across different models and sample formats, verifying the feasibility of the proposed air-writing trajectory framework for non-contact gesture-based UAV control. Furthermore, by combining gesture recognition with a GUI-based map planning interface, the system enhances the intuitiveness and convenience of UAV operation. Future extensions, such as incorporating aerial image object recognition, could extend the framework’s applications to scenarios including forest disaster management, vehicle license plate recognition, and air pollution monitoring.