The system, called FUS-DETR, was developed by Siyuan Duan, Geng Zhang, Xin Li, Shang Wang, Zhimin Mao, Bingliang Hu, and Xuebin Liu at the Xi’an Institute of Optics and Precision Mechanics of the Chinese Academy of Sciences, together with the University of Chinese Academy of Sciences. Writing in the journal Applied Intelligence, the team reports that their transformer-based detector achieves state-of-the-art results on two demanding air-to-air drone detection benchmarks, reaching 96.6 percent mean average precision at a 50 percent overlap threshold on the Det-Fly dataset and 98.1 percent on the UAV-Eagle dataset.

Those numbers matter because the air-to-air detection scenario is uniquely brutal for computer vision. When a camera drone is used to hunt another drone, the target often occupies only a handful of pixels in the frame. At long range, a quadcopter can shrink to a few dozen pixels across, losing nearly all of the distinctive shape cues that ground-based detectors rely on. Worse, the hunting drone’s own camera is moving, so the background — buildings, trees, roads, clouds, birds — streams past and creates a constantly shifting field of clutter that can masquerade as a target or swallow a real one entirely.

Traditional approaches to this problem have leaned heavily on convolutional neural networks, particularly the YOLO family of real-time detectors, which have dominated practical object detection for years. But convolutional networks process images through local receptive fields, which limits their ability to reason about the global context of a scene. A transformer-based detector, by contrast, uses self-attention mechanisms that let the model relate any part of the image to any other part, making it easier to distinguish a small intruder from its surroundings based on scene-wide patterns rather than purely local appearance.

FUS-DETR builds on the DETR, or Detection Transformer, architecture — specifically the real-time RT-DETR lineage that has recently begun to challenge YOLO models on speed and accuracy. The researchers introduced three key innovations. The first is a Feature-Focused Diffusion Pyramid Network, or FFDP, which tackles a chronic weakness of feature pyramid networks: as information flows between different scales of the image representation, fine details tend to be lost. The FFDP combines feature focus modules with a feature diffusion mechanism that enriches multi-scale feature maps with contextual information, deliberately spreading information across scales so that the faint signature of a small drone is not diluted or discarded along the way.

The second innovation is a Multilevel Attention Fusion Module, or MAF. Detection networks typically blend features from several layers, each capturing different levels of detail — shallow layers hold fine-grained edges and textures, while deeper layers encode broader semantic context. The MAF uses a hierarchical attention mechanism to adaptively fuse local and global features, learning on the fly which combination best captures the fine details of a small target while suppressing the background noise that would otherwise trigger false detections. In effect, the module lets the network decide, for every part of every image, how much to trust close-up texture versus wide-angle context.

The third component consists of three independent information fusion blocks arranged in concatenation and parallel structures. These blocks aggregate multi-scale information efficiently, giving the detection head a richer, more complete representation of each candidate target. Together, the three modules form a pipeline that is explicitly engineered around the two defining difficulties of drone detection from the air: extreme target smallness and heavy background clutter.

The experimental results are notable not just for their accuracy but for the rigor behind them. The team benchmarked FUS-DETR against a broad field of competitors, including classic detectors such as Faster R-CNN, SSD, and RetinaNet, the full YOLO series from YOLOv3 through YOLOv10, and the latest transformer-based detectors including RT-DETRv2 and DEIM. In an unusual display of transparency, the researchers disclosed in supplementary materials that they had initially reproduced some baseline results with an insufficient 50-epoch training schedule; they retrained RT-DETRv2 and DEIM under the full 100-epoch protocol and updated the comparison tables accordingly. All baselines were run under unified, fully documented training settings with the same 640 by 640 input resolution, ensuring that the comparison reflects the architectures rather than differences in training recipes.

The researchers also published detailed module-level complexity analyses, breaking down the parameter counts and computational costs of the FocusBlock and MAF components, including the multi-kernel depth-wise convolutions with kernel sizes of 5, 7, 9, and 11 that drive the feature diffusion process. This kind of accounting matters for a practical reason the authors highlight directly: achieving an optimal balance between top-tier detection performance and a model size suitable for aerial deployment remains a critical challenge. A detector that runs only on a data-center GPU is useless for a drone that must spot intruders in real time using onboard compute. The team’s attention to lightweight, re-parameterizable convolutional refinement operators suggests the design was shaped by deployment constraints from the start.

The datasets underpinning the results are publicly available — Det-Fly on GitHub and UAV-Eagle in an open repository — which means other groups can verify the numbers and build on them. The code is available from the corresponding author upon request. The work was supported by the Open Research Fund of the Shaanxi Key Laboratory of Optical Remote Sensing and Intelligent Information Processing, and the authors declare no competing financial interests.

If the reported performance holds up in independent testing, the implications extend well beyond counter-drone applications. The core techniques — context-preserving feature diffusion across pyramid scales, adaptive local-global attention fusion, and efficient multi-scale aggregation — address the general problem of small object detection in cluttered scenes, which also plagues medical imaging, satellite reconnaissance, wildlife monitoring, and autonomous driving. As drones multiply in the airspace, the ability of one flying machine to reliably see another may become as fundamental to aviation safety as radar was to the last century. FUS-DETR offers a glimpse of what that machine vision might look like: fast, compact, and sharp-eyed enough to find a speck of a drone in a sky full of noise.

Subject of Research: Transformer-based deep learning for detecting small unmanned aerial vehicles in cluttered airborne images

Article Title: FUS-DETR: a robust algorithm for drone detection from a cluttered background in airborne images

Article References: Duan, S., Zhang, G., Li, X., Wang, S., Mao, Z., Hu, B., & Liu, X. (2026). FUS-DETR: a robust algorithm for drone detection from a cluttered background in airborne images. Applied Intelligence, 56(15), Article 424. https://doi.org/10.1007/s10489-026-07448-y

Image Credits: AI Generated

DOI: 10.1007/s10489-026-07448-y

Keywords: UAV detection, FUS-DETR, transformer, object detection, small object detection, air-to-air detection, computer vision, feature pyramid network, attention mechanism, deep learning, drone safety, Applied Intelligence

The problem the team set out to address is known as actuator loss-of-effectiveness, or LoE. In a quadrotor, four rotors share the work of stabilizing the vehicle in roll, pitch, yaw, and altitude. If a single rotor degrades, whether through motor wear, a damaged propeller, or a partial power failure, the control authority available to the flight computer shrinks asymmetrically. A controller tuned for a healthy aircraft, with fixed gains calculated once before takeoff, may respond too weakly or too aggressively to the resulting imbalance, and the vehicle can destabilize. Fault-tolerant control strategies exist, ranging from robust backstepping designs to model predictive control and hardware redundancy such as tilting rotors, but many require substantial redesign of the control stack or additional actuators.

The Algerian team’s approach is deliberately conservative. They kept the standard cascaded PID structure, the underlying control mixer, and the rigid-body model of the quadrotor untouched. On top of that, they added a bounded online gain-adaptation layer implemented with reinforcement learning. Four decentralized agents operate in parallel, each responsible for one control channel: roll, pitch, yaw, and altitude. Each agent observes the state of the vehicle and adjusts the PID gains in its own channel in real time, but only within preset limits. This bounding is a critical safety feature, because it prevents the learning system from ever commanding gains that could make the aircraft unstable, a concern that has historically limited the acceptance of learning-based controllers in safety-critical flight applications.

To test the idea rigorously, the researchers built a six-degree-of-freedom Newton–Euler quadrotor model with first-order motor dynamics in MATLAB/Simulink. This level of modeling captures both the full rigid-body motion of the aircraft and the lag with which real motors respond to commands, which matters greatly when a rotor’s effectiveness is dropping. The adaptation layer was instantiated with Soft Actor-Critic, or SAC, a reinforcement learning algorithm known for balancing exploration and stability during training. The simulations subjected the controller to nominal flight conditions and to multiple transient and sustained single-rotor LoE profiles, meaning scenarios in which a rotor’s effectiveness dropped either briefly or permanently during flight.

A key strength of the study is its comparison set. The authors did not simply benchmark their learning controller against a naive baseline. They included two other prominent reinforcement learning algorithms, Twin Delayed Deep Deterministic Policy Gradient (TD3) and Proximal Policy Optimization (PPO), evaluated under identical observation and action definitions, the same reward structure, the same gain limits, and the same 500-episode training budget. An additional 1000-episode run of PPO was included to check whether the results were sensitive to how long the algorithms were allowed to train. This kind of controlled comparison is rare and valuable, because reinforcement learning results can vary dramatically with small changes in setup.

The team also addressed a subtler question: how much of the benefit comes from online adaptation itself, rather than from clever static tuning? To separate the two effects, they created offline-optimized fixed-gain PID baselines using two metaheuristic optimization methods, particle swarm optimization (PSO) and grey wolf optimization (GWO). These baselines represent the best that conventional, non-adaptive tuning can achieve before the flight even begins. Including two different metaheuristics also guards against the criticism that the comparison depends on which optimization algorithm happened to be chosen for the baseline.

The results tell a clear story. In nominal flight, with all rotors healthy, the methods performed comparably: the learning-augmented controller did not sacrifice accuracy in ordinary conditions, and step responses and three-dimensional trajectory-tracking simulations showed similar performance across the board. The differences emerged under degradation. When a rotor began to lose effectiveness, the fixed-gain controllers, even those tuned by sophisticated metaheuristics, showed larger post-fault deviations from the desired trajectory. The online gain adaptation layer, by contrast, produced smaller deviations after the fault, because the agents could shift the controller’s aggressiveness to compensate for the lost control authority as the degradation unfolded.

Among the three reinforcement learning algorithms tested, SAC provided the most consistent fault accommodation in the evaluated configurations. This finding aligns with SAC’s design philosophy: the algorithm optimizes both the expected reward and the entropy of its policy, encouraging robust behavior rather than overfitting to a narrow set of training conditions. TD3 and PPO remained competitive under the same budget, but SAC’s consistency across the various LoE profiles made it the standout. The additional 1000-episode PPO check helped the authors assess whether longer training would change the picture, addressing a common concern that reinforcement learning comparisons may simply reflect training-budget artifacts.

What makes this work notable for the drone industry is its integration cost, or rather its lack of one. Many fault-tolerant control approaches demand that engineers abandon the PID controllers their teams know well and adopt entirely new architectures, with all the certification, testing, and retraining burdens that implies. The approach demonstrated here treats learning as a thin adaptation layer on top of existing infrastructure. The mixer, the outer-loop structure, and the physical model all remain unchanged, and the learning agents act only within preset gain bounds. For operators of delivery drones, inspection platforms, and other commercial quadrotors, that means a path to greater resilience against rotor degradation without rewriting the flight stack from scratch.

The study is simulation-based, and the authors are careful about the scope of their claims: in the tested setup, the results support bounded online gain adaptation as a low-integration-cost way to improve robustness to rotor loss-of-effectiveness. Real-world deployment would bring additional challenges, including sensor noise, wind, computational constraints on embedded flight controllers, and the well-known sim-to-real gap that affects all learning-based control methods. Still, the work adds to a growing body of evidence that reinforcement learning can serve flight control best not as a wholesale replacement for classical control theory, but as a disciplined assistant that fine-tunes proven controllers when the aircraft’s condition changes. As drones take on ever more demanding missions, that kind of graceful degradation under failure may prove to be one of machine learning’s most practical contributions to aviation safety.

Subject of Research: Reinforcement-learning-based online PID gain adaptation for fault-tolerant quadrotor control under rotor loss-of-effectiveness

Article Title: Reinforcement-Learning-Based Online PID Gain Adaptation for Fault-Tolerant Quadrotor Control Under Rotor Loss-of-Effectiveness

Article References: Lahmar, O., Abdou, L., & Ghiloubi, I. B. (2026). Reinforcement-Learning-Based Online PID Gain Adaptation for Fault-Tolerant Quadrotor Control Under Rotor Loss-of-Effectiveness. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01269-6

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01269-6

Keywords: quadrotor UAV, fault-tolerant control, actuator loss-of-effectiveness, reinforcement learning, PID gain adaptation, Soft Actor-Critic, TD3, PPO, MATLAB/Simulink simulation, trajectory tracking, drone safety, adaptive control

The research, carried out by Neha M V and Sabu M Thampi at the Digital University Kerala’s School of Computer Science and Engineering, tackles two intertwined problems that have long limited urban drone autonomy. The first is perception: dynamic obstacles such as vehicles, pedestrians and other aircraft move unpredictably, and detecting them in time requires processing enormous streams of video, radar and other sensor data. The second is security: GPS spoofing, in which an adversary broadcasts counterfeit satellite signals, can steer a drone off course or into danger. Current approaches usually address these problems separately, and the frameworks that do combine them tend to be computationally heavy, making real-time, interpretable decision-making under uncertainty difficult on the small processors a UAV can actually carry.

The centrepiece of the new framework is a two-stage defensive and navigational architecture. The first stage is dedicated to trust: a sparse autoencoder, a neural network trained to reconstruct the statistical fingerprints of genuine GPS signals, continuously monitors incoming navigation data. Sparse autoencoders work by compressing inputs through a bottleneck layer while imposing sparsity constraints, so they learn only the essential structure of legitimate signals. When a spoofed signal arrives, the reconstruction error spikes, flagging an anomaly the drone can act on. This detection module acts as a gatekeeper; as long as GPS readings look normal, the system operates conventionally, but the moment an anomaly is detected, the platform shifts into a degraded-GPS mode where navigation integrity is maintained through other means.

That shift is where the second stage comes in. When GPS performance degrades, the framework engages a multi-sensor fusion process that blends information from complementary sources, including vision-based detection, radar and other onboard sensing modalities. The philosophy behind sensor fusion is straightforward in principle and demanding in practice: each sensor has blind spots and failure modes, but their errors are largely uncorrelated, so combining them produces a more reliable picture of the environment than any single instrument could. Cameras offer rich visual detail but struggle in low light; radar penetrates fog and darkness but provides coarse spatial resolution. Fusing their outputs, with tracking stages informed by techniques such as extended Kalman filtering, allows the drone to detect, locate and track moving obstacles even when one channel of information is compromised. Crucially, the researchers designed this fusion pipeline to maximise computational efficiency rather than to throw raw processing power at the problem.

The efficiency gains come in large part from keyframe extraction. Video streams aboard a UAV contain enormous redundancy, with consecutive frames differing only slightly. Rather than pushing every frame through computationally expensive perception models, the system selects informative keyframes that capture the essential changes in the scene and analyses those. This strategy alone reduces the inference load by a striking factor of 122.5, which is what makes the pipeline feasible for real-time operation on resource-constrained aerial hardware. For a drone dodging a delivery drone head-on or tracing a vehicle through dense traffic, milliseconds matter, and shaving the computational burden of perception is not a luxury but a precondition for safety.

Detection, however, is only half of the autonomy problem. Once the drone knows where the hazards are, it must decide what to do about them, and the researchers argue that black-box neural networks are poorly suited to that role in safety-critical flight. Their answer is to embed Logical Neural Networks, or LNNs, into the decision-making core. LNNs are a hybrid form of artificial intelligence that represents logical rules inside neural architectures, so that reasoning is both learnable from data and traceable in human-readable form. Instead of an opaque model simply outputting an avoidance command, an LNN can offer context-aware decisions whose basis, the obstacles detected, the navigation state, and the rules governing safe flight, can be inspected and audited. This interpretability matters for regulators, for engineers debugging flight behaviour, and for any operator who must eventually explain to an accident investigator why a drone did what it did.

The team benchmarked the framework against interpretable baseline systems on publicly available datasets, drawing on urban sensor data and UAV-specific resources that include the GREAT Dataset of vehicle-mounted multi-sensor observations in complex city environments, the VisDrone object detection collection, the MAN TruckScenes multimodal dataset, and the IEEE DataPort UAV attack dataset. Across those evaluations, the combined system achieved an overall accuracy of 90 percent with a 90 percent F1-score, and, notably, a 75 percent emergency recall, meaning it correctly identified three-quarters of emergency situations requiring avoidance action. The authors report that these figures outperform other interpretable baselines while simultaneously reducing inference load through the keyframe strategy, a combination they argue establishes meaningful improvements in navigation integrity, system robustness and decision transparency.

The significance of the work lies partly in what it refuses to trade away. Plenty of machine learning systems can match or beat 90 percent accuracy on a benchmark, but far fewer can do so while explaining their reasoning, while running on the fly, and while remaining resilient to deliberate adversarial interference. GPS spoofing is no longer a hypothetical threat; the researcher community has documented attacks against civilian drones, and the specter of a hijacked UAV crashing into a crowd or critical infrastructure has pushed anti-spoofing techniques, including support vector machine-based detection methods and sparse autoencoder-based anomaly detection, into the mainstream of aerial robotics research. By tying spoofing detection directly into a fallback navigation strategy, the new framework treats security and safety as a single continuous problem rather than two separate engineering silos.

There are, of course, limitations inherent to any experimental evaluation, and the benchmarks used here, however diverse, cannot fully reproduce the chaos of a real metropolitan sky with its rain, magnetic interference, RF congestion and unpredictable human behaviour. The authors themselves frame the contribution as establishing a foundation: a fusion-based navigation architecture that stays interpretable and computationally light enough for deployment. The funding came through a fellowship from the Kerala University of Digital Sciences, Innovation and Technology, and the work reflects a broader movement toward trustworthy autonomy, where explainable reasoning engines like LNNs and anomaly-detection components like sparse autoencoders are woven together rather than bolted on after the fact.

If the vision holds up in field trials, the implications stretch well beyond the drone itself. The same recipe, anomaly detection at the signal level, multi-modal sensor fusion at the perception level, and logical neural reasoning at the decision level, could apply to self-driving cars, warehouse robots and any machine expected to make safety-critical choices in a world that sometimes lies to it. For now, the study offers a concrete demonstration that a drone can be made to notice when its compass of the world is being forged, switch to its own senses, and still find its way home with the reasons for every swerve written down in a form a human can read. In an era when autonomous machines are being asked to share increasingly crowded airspace, that combination of robustness and transparency may prove to be the most important flight instrument of all.

Subject of Research: Autonomous UAV obstacle avoidance using multi-sensor fusion and interpretable decision-making against GPS spoofing

Article Title: A robust autonomous UAV obstacle avoidance through multi-sensor fusion and intelligent decision-making

Article References: M V, N., & Thampi, S. M. (2026). A robust autonomous UAV obstacle avoidance through multi-sensor fusion and intelligent decision-making. Multimedia Tools and Applications, 85(10), Article 768. https://doi.org/10.1007/s11042-026-21913-3

Image Credits: AI Generated

DOI: 10.1007/s11042-026-21913-3

Keywords: UAV, obstacle avoidance, multi-sensor fusion, GPS spoofing, sparse autoencoder, Logical Neural Networks, autonomous navigation, explainable AI, keyframe extraction, drone safety, smart cities, real-time decision-making