Luis F. F. M. Santos, Miguel Ángel Sánchez-Tena, Cristina Alvarez-Peregrina
et al.
This study introduces an Artificial Intelligence framework based on the Deep Learning model Bidirectional Encoder Representations from Transformers framework trained on a dataset from 2000–2023. The AI tool categorizes articles into six classes: Contactology, Low Vision, Refractive Surgery, Pediatrics, Myopia, and Dry Eye, with supervised learning enhancing classification accuracy, achieving F1-Scores averaging 86.4%, AUC at 0.98, Precision at 87%, and Accuracy at 86.8% via one-shot training, while Epoch training showed 85.9% Accuracy and 92.8% Precision. Utilizing the Artificial Intelligence model outputs, the Autoregressive Integrated Moving Average model provides forecasts from all classes through 2030, predicting decreases in research interest for Contactology, Low Vision, and Refractive Surgery but increases for Myopia and Dry Eye due to rising prevalence and lifestyle changes. Stability is expected in pediatric research, highlighting its focus on early detection and intervention. This study demonstrates the effectiveness of AI in enhancing diagnostic precision and strategic planning in optometry, with potential implications for broader clinical applications and improved accessibility to eye care.
For the non-monotone Hamilton-Jacobi equations of contact type, the associated Lax-Oleinik semiflow ( T t , C ( M ) ) (T_t, C(M)) is expansive. In this paper, we provide qualitative estimates for both the lower and upper bounds of the topological entropy of the semiflow.
Cannabis is a strictly regulated plant in China, and its illegal cultivation presents significant challenges for social governance. Traditional manual patrol methods suffer from low coverage efficiency, while satellite imagery struggles to identify illicit plantations due to its limited spatial resolution, particularly for sparsely distributed and concealed cultivation. UAV remote sensing technology, with its high resolution and mobility, provides a promising solution for cannabis monitoring. However, existing detection methods still face challenges in terms of accuracy and robustness, particularly due to varying target scales, severe occlusion, and background interference. In this paper, we propose AMS-YOLO, a cannabis detection model tailored for UAV imagery. The model incorporates an asymmetric backbone network to improve texture perception by directing the model’s focus towards directional information. Additionally, it features a multi-scale fusion neck structure, incorporating partial convolution mechanisms to effectively improve cannabis detection in small target and complex background scenarios. To evaluate the model’s performance, we constructed a cannabis remote sensing dataset consisting of 1972 images. Experimental results show that AMS-YOLO achieves an mAP of 90.7% while maintaining efficient inference speed, outperforming existing state-of-the-art detection algorithms. This method demonstrates strong adaptability and practicality in complex environments, offering robust technical support for monitoring illegal cannabis cultivation.
This paper focuses on the study of uniform demagnetization fault diagnosis in permanent magnet synchronous motors （PMSMs） and proposes a novel method based on current signals for diagnosing faults at different degrees of uniform demagnetization. An improved ensemble empirical mode decomposition （EEMD） algorithm combined with particle swarm optimization‑support vector machine （PSO‑SVM） is introduced for fault diagnosis. First， the improved EEMD is employed to denoise and reconstruct the collected stator current signals. Then， the fractal box dimension of the processed data is calculated as the fault feature parameter. Finally， PSO‑SVM is utilized to diagnose uniform demagnetization faults based on the extracted feature parameters. Simulation experiments and prototype testing demonstrate that the proposed method accurately identifies uniform demagnetization faults in PMSMs， achieving an average recognition rate of over 96%， thus validating the effectiveness of the proposed fault diagnosis approach.
Jorge Luis Alva Alarcon, Yan Rockee Zhang, Hernan Suarez
et al.
The increasing demand for noninvasive inspection (NII) of complex civil infrastructures requires overcoming the limitations of traditional ground-penetrating radar (GPR) systems in addressing diverse and large-scale applications. The solution proposed in this study focuses on an initial design that integrates a low-SWaP (Size, Weight, and Power) ultra-wideband (UWB) impulse radar with realistic electromagnetic modeling for deployment on unmanned aerial vehicles (UAVs). The system incorporates ultra-realistic antenna and propagation models, utilizing Finite Difference Time Domain (FDTD) solvers and multilayered media, to replicate realistic airborne sensing geometries. Verification and calibration are performed by comparing simulation outputs with laboratory measurements using varied material samples and target models. Custom signal processing algorithms are developed to extract meaningful features from complex electromagnetic environments and support anomaly detection. Additionally, machine learning (ML) techniques are trained on synthetic data to automate the identification of structural characteristics. The results demonstrate accurate agreement between simulations and measurements, as well as the potential for deploying this design in flight tests within realistic environments featuring complex electromagnetic interference.
This paper presents the observations of temporally evolving stochastic vibration patterns of a coin vibrating motor. Various voltages are applied to the coin vibrating motor, and the resulting vibrations are recorded using an accelerometer. Although an overall upward trend in mean vibration amplitude is observed with increasing drive voltage, instantaneous waveforms displayed pronounced nonlinear and quasiperiodic amplitude modulations, frequency shifts, and stochastic deviations that intensified at higher voltages. Additional experiments involving periodic pressing of the motor and the propagating medium revealed the dependence of nonlinear electromechanical responses on the initial conditions. These results demonstrate that the dynamic behaviour of the coin-type motor is governed by a complex nonlinear dependence on current and displacement, with significant implications for precision control in miniature actuator applications.
Modelling pedestrian-driver interactions is critical for understanding human road user behaviour and developing safe autonomous vehicle systems. Existing approaches often rely on rule-based logic, game-theoretic models, or 'black-box' machine learning methods. However, these models typically lack flexibility or overlook the underlying mechanisms, such as sensory and motor constraints, which shape how pedestrians and drivers perceive and act in interactive scenarios. In this study, we propose a multi-agent reinforcement learning (RL) framework that integrates both visual and motor constraints of pedestrian and driver agents. Using a real-world dataset from an unsignalised pedestrian crossing, we evaluate four model variants, one without constraints, two with either motor or visual constraints, and one with both, across behavioural metrics of interaction realism. Results show that the combined model with both visual and motor constraints performs best. Motor constraints lead to smoother movements that resemble human speed adjustments during crossing interactions. The addition of visual constraints introduces perceptual uncertainty and field-of-view limitations, leading the agents to exhibit more cautious and variable behaviour, such as less abrupt deceleration. In this data-limited setting, our model outperforms a supervised behavioural cloning model, demonstrating that our approach can be effective without large training datasets. Finally, our framework accounts for individual differences by modelling parameters controlling the human constraints as population-level distributions, a perspective that has not been explored in previous work on pedestrian-vehicle interaction modelling. Overall, our work demonstrates that multi-agent RL with human constraints is a promising modelling approach for simulating realistic road user interactions.
This study investigates the issue of multi-objective mission planning for multi-payload satellite constellations via the nondominated sorting carnivorous plant algorithm (NSCPA). Observation time windows are generated, and a constraint satisfaction model is established based on multiple regional targets, satellite orbits, and characteristics of the synthetic aperture radar (SAR) payload and optical payload. A task conflict detection and resolution method is proposed to handle the task assignment among multiple satellites. Based on the existing single objective-based CPAs, a modified multi-objective NSCPA is first developed for multi-objective planning optimization using the non-dominated sorting algorithm. The effectiveness and superiority of the NSCPA are verified by a series of simulation experiments and comparisons with the traditional non-dominated sorting genetic algorithms-II (NSGA-II) and particle swarm optimization (PSO).
Technology, Motor vehicles. Aeronautics. Astronautics
Directional unmanned aerial vehicle (UAV) ad hoc networks (DUANETs) are widely applied due to their high flexibility, strong anti-interference capability, and high transmission rates. However, within directional networks, complex mutual interference persists, necessitating scheduling of the time slot, power, and main lobe direction for all links to improve the transmission performance of DUANETs. To ensure transmission fairness and the total count of transmitted data packets for the DUANET under dynamic data transmission demands, a scheduling algorithm for the time slot, power, and main lobe direction based on multi-agent deep reinforcement learning (MADRL) is proposed. Specifically, modeling is performed with the links as the core, optimizing the time slot, power, and main lobe direction variables for the fairness-weighted count of transmitted data packets. A decentralized partially observable Markov decision process (Dec-POMDP) is constructed for the problem. To process the observation in Dec-POMDP, an attention mechanism-based observation processing method is proposed to extract observation features of UAVs and their neighbors within the main lobe range, enhancing algorithm performance. The proposed Dec-POMDP and MADRL algorithms enable distributed autonomous decision-making for the resource scheduling of time slots, power, and main lobe directions. Finally, the simulation and analysis are primarily focused on the performance of the proposed algorithm and existing algorithms across varying data packet generation rates, different main lobe gains, and varying main lobe widths. The simulation results show that the proposed attention mechanism-based MADRL algorithm enhances the performance of the MADRL algorithm by 22.17%. The algorithm with the main lobe direction scheduling improves performance by 67.06% compared to the algorithm without the main lobe direction scheduling.
Lü Zhenrui, Shen Xin, Li Shaobo, Tian Peng, Si Yingli
Facing the increasing complexity of aerial combat environments and challenges to the survivability of air platforms from new combat methods, it is necessary to adopt new hard-kill methods to counter advanced air-to-air missiles. In order to improve the success rate and efficiency of launching air-to-air missiles to intercept incoming missiles as a hard kill method, this study proposes intelligent maneuvering strategies for aircraft platforms and missile interception strategies based on reinforcement learning. Firstly, this paper designs the missile threat assessment technology, constructs the simulation environments, and determines the strategy model state and reward function. By setting various attack angles and positions of incoming air-to-air missiles and training maneuvering and intelligent interception strategies under different aircraft platform postures, this paper achieves active interception of incoming targets and effective maneuvering of the aircraft platform. Experiments show that compared to the average escape probability of 5.8% in operations research game strategies, after using maneuver and interception strategies based on reinforcement learning, the average escape probability can increase to 56.8%; Meanwhile, the utilization rate of interceptors has increased by approximately 13.3%, and the response time has remained within 24 ms. The designed strategy can adapt to different numbers of incoming missiles, can significantly improve the survival ability of the carrier platform and the success rate of intercepting incoming missiles. This study can support continuous optimization in a high-dimensional state space of air combat.
The problem of selecting suboptimal parameters of an electric propulsion system as part of a spacecraft, ensuring a geostationary orbit raising mission using a Soyuz-5 launch vehicle at the initial launch stage, is investigated. An expression is obtained for the main optimality criterion – the payload relative mass. An algorithm is obtained for solving the problem of parametric optimization of an electric propulsion system as part of a hybrid transportation system. The task of selecting the optimal electric propulsion engine for launching a spacecraft with an electric propulsion system to a geostationary orbit using the launch vehicle under consideration is solved. A design ballistic calculation and a comparative analysis of the obtained data are performed.
Gabriele Berton, Alex Stoken, Barbara Caputo
et al.
Astronaut photography, spanning six decades of human spaceflight, presents a unique Earth observations dataset with immense value for both scientific research and disaster response. Despite its significance, accurately localizing the geographical extent of these images, crucial for effective utilization, poses substantial challenges. Current manual localization efforts are time-consuming, motivating the need for automated solutions. We propose a novel approach - leveraging image retrieval - to address this challenge efficiently. We introduce innovative training techniques, including Year-Wise Data Augmentation and a Neutral-Aware Multi-Similarity Loss, which contribute to the development of a high-performance model, EarthLoc. We develop six evaluation datasets and perform a comprehensive benchmark comparing EarthLoc to existing methods, showcasing its superior efficiency and accuracy. Our approach marks a significant advancement in automating the localization of astronaut photography, which will help bridge a critical gap in Earth observations data. Code and datasets are available at https://github.com/gmberton/EarthLoc
Marco P. M. de Souza, Sidnei P. Oliveira, Valdenice L. Luiz
In this work, we present the Electric Motor simulator, an application from the SimuFísica\textsuperscript{\textregistered} platform designed for classroom use. We briefly describe the technologies behind the application, the equations that govern its operation, some studies showing the dynamics of the electric motor, and, finally, the use of the application in High School and Higher Education.
Aerodynamic characteristics of revolving wing models were investigated to assess the validity of the normal force model. Aerodynamic force and torque measurements were conducted for six wing planforms (with aspect ratios of 2 and 3, and area centroid locations at 40%, 50%, and 60% of the wing length) at three different Reynolds numbers (0.5 × 10<sup>4</sup>, 1 × 10<sup>4</sup>, and 1.5 × 10<sup>4</sup>) and three thickness-to-chord ratios (3%, 4%, and 5%). Both early and steady phase measurements were extracted for a range of angles of attack relevant to insect flight. It was shown that the so-called “normal force” model conveniently captures the variation of the lift and drag coefficients along the first quadrant of angles of attack for all cases tested. A least squares best fit model for the obtained experimental measurements was used to estimate the key parameters of the normal force model, namely the lift curve slope, the zero-lift drag coefficient, and the peak drag coefficient. It was shown that the knowledge of only the lift curve slope and the zero-lift drag coefficient is sufficient to fully describe the model, and that clear trends of these two parameters exist. Notably, both parameters decreased with the increase in area centroid location. For instance, for steady measurements and on average, the lift curve slope for a wing with an area centroid location at 40% span was 15.6% higher compared to an area centroid location at 60% span. However, the increase in the zero-lift drag coefficient for wings with a lower area centroid location had a detrimental effect on aerodynamic efficiency assessed via glide ratio. Wings with a lower area centroid location consistently led to a lower glide ratio regardless of the change in aspect ratio, thickness-to-chord ratio, or Reynolds number. Increasing the aspect ratio decreased the zero-lift drag coefficient but generally had a slighter increasing effect on the lift curve slope. Increasing the Reynolds number within the range experimented decreased both the lift curve slope and the zero-lift drag coefficient. Finally, the effect of the thickness-to-chord ratio was mainly pronounced in its effect on the zero-lift drag coefficient.
In order to overcome the influence of complex environmental disturbance factors such as nonlinear time-varying characteristics on the dynamic control performance of small fixed-wing UAVs, the nonlinear expression relationship of neural networks (NNs) is combined with the recursive least squares (RLSs) identification algorithm. This paper proposes a hybrid aerodynamic parameter identification method based on NN-RLS offline network training and online learning correction. The simulation results show that compared with the real value of the identification value obtained by this algorithm, the residual error of the moment coefficient is reduced by 69%, and the residual error of the force coefficient is reduced by 89%. Under the same identification accuracy, the identification time is shortened from the original 0.1 s to 0.01 s. Compared with traditional identification algorithms, better estimation results can be obtained. By using this algorithm to continuously update the NN model and iterate repeatedly, iterative learning for complex dynamic models can be realized, providing support for the optimization of UAV control schemes.