This paper investigates the model-free tracking control problem for unmanned surface vehicles (USVs) characterized by highly nonlinear dynamics, strong coupling, and significant uncertainties. A new model-free optimization backstepping (MFOB) tracking controller is proposed for USVs tracking with unknown dynamics. A second-order model is adopted to describe the USVs' dynamics. Within the backstepping framework, the USV tracking system is decomposed into two interconnected subsystems. By optimizing the control strategy for each subsystem individually, the overall system performance is enhanced, thereby achieving global control optimization. To present optimal performance control scheme for each subsystem, a new Hamilton-Jacobi-Bellman (HJB) equation is formulated with respect to the tracking error and control input within each subsystem. The HJB equation is highly nonlinear and complex, hence the actor-critic deep reinforcement learning method is chosen to search for the optimal or suboptimal solution of the HJB equation. In actor-critic methods, performance is often critically dependent on an accurate representation of the underlying system dynamics. To address this issue, an extended state observer incorporating fractional power functions is employed to design an online real-time identifier for the model. Incorporating model identification can effectively mitigate the actor-critic algorithm's dependence on a precise system model. The efficacy of the MFOB is verified through the numerical simulations.

Accurate prediction of energy consumption is essential for the operation and charging management of battery-electric buses. Existing prediction studies are often constrained by incomplete or low-resolution input data, limiting their robustness under real-world operating conditions. This paper presents a high-resolution, sensor-rich energy consumption modeling framework using second-by-second operational data and tests on an electric bus fleet operating on Route 49 in Jinan, China. The dataset integrates synchronized measurements of vehicle kinematics, powertrain variables, and thermal conditions, providing a substantially more complete description of bus operation against previous studies. Boosting-based machine learning models are developed to predict the instantaneous power demand, and their performance is evaluated in comparison with a physics-based energy model and other variants of machine learning models. Results show that the data-driven boosting models demonstrate excellent explanatory power (R2 values of up to 0.99 (training) and 0.95 (test)) and remain reliable under nonlinear operating conditions. Feature and SHAP analyses identify physically consistent energy drivers, supporting the applicability of the approach to real-world public transport operations.

Pseudolite (PL) positioning technology is one of the effective methods to achieve high-precision indoor positioning. The Distributed Array Pseudolite System (DAPLS) is a ground-based augmentation architecture designed to provide high-precision positioning in GNSS-denied or indoor environments. However, maintaining the stability and integrity of pseudolite signals in distributed deployments remains a significant challenge. To address this, a Pseudolite Monitoring Station (PMS) was developed for real-time signal observation, performance evaluation, and anomaly detection. The proposed PMS integrates a multi-channel front-end, signal-processing engine, and monitoring algorithms capable of continuous assessment across three hierarchical levels: Signal Quality Monitoring (SQM), Receiver Processing Monitoring (RPM), and Measurement Quality Monitoring (MQM). To integrate multi-domain monitoring results, a Composite Quality Index (CQI) model is introduced, combining normalized sub-scores through weighted fusion to reflect overall system integrity. A comprehensive Signal Quality Assessment (SQA) framework is further introduced, including four dimensions of evaluation: constellation status, time reference, spatial coordinate reference, and signal anomaly detection. An indoor DAPLS experiment was conducted within a laboratory-level test field. The system comprised three pseudolite transmitter arrays (six transmitters each) and a central monitoring station. Experimental results showed stable synchronization within ±5 ns, coordinate accuracy within 0.2 m, and consistently high signal quality. The monitoring station effectively detected minor signal distortions and synchronization deviations, confirming its diagnostic precision and robustness. This study demonstrates a complete monitoring and evaluation framework for DAPLS, enabling both system-level quality assurance and signal integrity monitoring. The proposed PMS and SQA methods provide essential tools for future deployment of pseudolite-based indoor positioning and timing systems.

Real-time deployment of synthetic aperture radar (SAR) ship detection systems encounters substantial obstacles stemming from inherent speckle noise characteristics, intricate background interference patterns, and excessive computational demands in contemporary deep learning architectures. We present an adaptive multiscale feature-fusion SAR detection network (AMFS-Net), a streamlined framework that realizes efficient SAR vessel identification through three synergistic technological innovations. The C3k2_FasterWConv module leverages weighted convolution operations with selective quarter-channel processing, exploiting spatial density functions for dynamic weight adjustment while substantially diminishing computational complexity. A dual-scale efficient detection (DSED) head implements a P3+ P4 architecture featuring intelligent parameter sharing mechanisms, removing the computation-intensive P2 layer and yielding a 33% reduction in detection overhead. The adaptive parameter scaling optimization (APSO) framework executes joint depth-width-channel scaling coupled with sensitivity-guided preservation of critical layers, achieving 27%-34% parameter reduction. Furthermore, a hierarchical adaptive geometric optimization IoU (HAGOIoU) loss function employing quality-adaptive weighting effectively addresses speckle noise and sea clutter interference. Extensive experimental validation on SSDD and HRSID benchmark datasets demonstrates that AMFS-Net attains 98.2% and 93.1% mAP50, respectively, while reducing parameters by 38.4% and accelerating inference speed by 48.0% relative to the baseline YOLOv11n architecture. The proposed framework establishes an effective paradigm for real-time SAR ship detection deployment in resource-constrained computational environments while preserving superior detection accuracy.

High-precision, reliability, and stability in location based services are essential prerequisites for the automated operations of unmanned ground vehicles (UGV). As a competitive technology in indoor positioning applications, the ultra-wideband (UWB) can be integrated with INS for navigation and positioning of UGV in challenging environments such as indoor or urban underground spaces. In order to mitigate the adverse effect of model perturbation and measurement outliers on the performance of UWB/INS tightly coupled integration system in practical applications, a novel robust cubature Kalman filter (CKF) method based on the variational Bayesian approach and Sage Husa noise estimator is proposed. The proposed method enhances the filtering precision of the UWB/INS integrated system while enhancing both adaptability and robustness of state estimation. In addition, the external height constraint is imposed on the UWB/INS tightly coupled integration system to enhance its performance further. Real-world experimental results indicate that compared with CKF and robust CKF, the proposed method increases horizontal positioning accuracy by 27.81% and 25.34% in experiment I, and by 40% and 31.37% in experiment II, respectively. The proposed method can simultaneously mitigate the effect of inaccurate noise statistics and measurement outliers, thereby improving the adaptability and robustness of filtering estimation, and providing accurate state estimation.

With the rapid development of global tourism, traditional scenic area guidance equipment faces significant challenges when deployed in low-light night-time environments. Low brightness, low contrast, and noise interference severely degrade image quality, thereby reducing visitor safety and the efficiency of scenic area management. To tackle these challenges, this paper proposes a pedestrian detection and tracking method tailored for low-light conditions. By integrating image processing and deep learning technologies, it enhances the perception capabilities of conventional equipment in nocturnal scenarios. This approach employs a low-light enhancement module to adjust brightness and contrast in captured images, thereby improving visibility. A trajectory tracking module enables the monitoring of pedestrian movements during night-time hours. Through a joint training framework, an effective feedback mechanism is established between sub-networks, further enhancing detection and tracking accuracy. Experimental results demonstrate that the proposed solution significantly improves visitor detection and trajectory tracking performance in low-light scenarios, effectively mitigating the false and missed detections commonly encountered by conventional guided tour systems.

This paper addresses the challenges in low-light image enhancement, including weak global modeling capabilities and the difficulty in coordinating computational efficiency with detail preservation. We propose a fusion framework based on the Visual State Space Model (VSSM) and frequency-domain dynamic enhancement. By establishing a cross-domain collaborative mechanism, we innovatively design a dark-channel prior-guided parallel feature extraction and an illumination-reflection decoupling reconstruction strategy to achieve a physically interpretable enhancement process. The core algorithm consists of an Efficient Visual Scanning module (EVS) and a Discriminative Frequency-Domain Feedforward Network (EDFFN). The EVS module uses dynamic geometric transformation and a noise-aware scanning mechanism to model cross-regional spatial correlations while maintaining linear complexity. The EDFFN module achieves a dynamic balance between high-frequency detail enhancement and noise suppression through time-frequency dual-domain processing paths and adaptive band-pass filtering techniques. These two modules form a closed-loop optimization system of "long-range spatial modeling - local frequency-domain enhancement" through luminance gating fusion, improving model transparency and breaking through the performance bottlenecks of traditional methods.Experimental results show that our method significantly outperforms mainstream algorithms in terms of image quality, dynamic range expansion, and texture detail preservation. It effectively balances noise suppression and naturalness preservation, and recovers details in dark areas and preserves colors in extreme low-light scenes. Its high computational efficiency enables real-time processing of high-resolution images, providing an innovative solution for real-time visual enhancement scenarios such as intelligent security and autonomous driving.

This paper proposes an estimator design approach for the sideslip angle of unmanned surface vessels (USVs) based on an extended state observer with fractional power function. Here it is called as the ESO-fal. Based on the estimation of the sideslip angle, a new line-of-sight guidance law (EFLOS) is designed for a class of underactuated USV. The aim is to address the problem of rapidly time-varying sideslip angle tracking for underactuated USVs in complex marine environments. The ESO-fal is developed for rapid and accurate identification of the time-varying sideslip angle caused by marine dynamic disturbances. Under the influence of wind, waves, and currents in complex marine environments, disturbances substantially compromise path-following performance, underscoring the need to fundamentally enhance its accuracy and stability. Benefiting from the performance characteristics of the ESO-fal, the EFLOS guidance law has a faster convergence speed and better robustness. It is shown from simulation experiments that the EFLOS guidance law has more stable tracking effects, faster convergence speed and more accurate sideslip angle estimation under different speed cases.

Accurate and reliable positioning is a crucial foundation for achieving sufficient autonomy and intelligence in positioning systems. The ultra-wideband (UWB) technology has significant potential for applications in the challenging and complex field of indoor positioning, and is often integrated with the inertial navigation system (INS) to enhance the overall performance of integrated positioning system. This paper proposes a novel variational Bayesian-based cubature information filter that enhances the positioning accuracy of tightly coupled UWB/INS navigation system. This approach aims to provide continuous and reliable navigation and positioning across various scenarios. The proposed integrated system applies the non-holonomic constraint to positioning in the elevation direction, thereby enhancing the system's overall positioning accuracy and performance. Real-world experimental results demonstrate that the proposed system delivers continuous, accurate, and stable positioning. In UWB line-of-sight (LOS) scenario, the proposed system achieves root mean square errors (RMSEs) of 0.085 m, 0.091 m, and 0.125 m in the east, north, and horizontal directions, respectively; the corresponding errors in non-LOS (NLOS) scenario are 0.286 m, 0.237 m, and 0.372 m. Furthermore, when integrated with a scheme for mitigating NLOS errors, the proposed system demonstrates a significant enhancement in positioning accuracy. In conclusion, the experimental results effectively validate the proposed system's effectiveness in achieving accurate and continuous positioning across various environments, demonstrating its application potential in complex scenarios.

In high-precision fields such as advanced manufacturing, semiconductor processing, aerospace assembly, and precision machining, motion control systems often face challenges such as large tracking errors and low control efficiency due to complex dynamic environments. To address this, this paper innovatively proposes a data-driven feedforward compensation control strategy based on a Parallel Gated Recurrent Unit (GRU)–Transformer. This method does not require an accurate model of the controlled object but instead uses motion error data and controller output data collected from actual operating conditions to complete network training and real-time prediction, thereby reducing data requirements. The proposed feedforward control strategy consists of three main parts: first, a Parallel GRU–Transformer prediction model is constructed using real-world data collected from high-precision sensors, enabling precise prediction of system motion errors after a single training session; second, a nonlinear PD controller is introduced, using the prediction errors output by the Parallel GRU–Transformer network as input to generate the primary correction force, thereby significantly reducing reliance on the main controller; and finally, the output of the nonlinear PD controller is combined with the output of the main controller to jointly drive the precision motion platform. Verification on a permanent magnet synchronous linear motor motion platform demonstrates that the control strategy integrating Parallel GRU–Transformer feedforward compensation significantly reduces the tracking error and fluctuations under different trajectories while minimizing moving average (MA) and moving standard deviation (MSD), enhancing the system’s robustness against environmental disturbances and effectively alleviating the load on the main controller. The proposed method provides innovative insights and reliable guarantees for the widespread application of precision motion control in industrial and research fields.

To address the issue that existing image dehazing methods using convolutional neural networks (CNNs) as the backbone fail to utilize the deep detail features of hazy images, leading to suboptimal dehazing performance of trained models, this paper innovatively integrates network structure improvement and a two-stage feature fusion strategy based on a single-image dehazing method using Detail-Enhanced Convolution (DEConv) and Content-Guided Attention (CGA). For network architecture optimization, the improvement involves serially fusing the Spatial Pyramid Attention (SPA) module with the Detail Enhancement Attention Block (DEAB) to construct a hierarchical feature extraction system. The two-stage feature fusion establishes a cascade mechanism of cross-level attention mixing and multi-scale attention enhancement. These methods enhance the model's ability to capture multi-scale structural information and accurately perceive haze concentration distributions.

This paper proposes a feedforward compensation strategy based on Parallel GRU-Transformer neural network to address the issues of large tracking errors and insufficient stability of multi degree of freedom precision motion platforms in complex dynamic environments. By accurately predicting the output of the feedback controller at the next moment and using the predicted results as feedforward compensation inputs, the overall performance of the control system is significantly improved. First, a Parallel GRU-Transformer neural network was constructed to compensate for the limitations of LSTM and GRU neural networks in parallel processing due to their recursive structure, enabling the network to accurately predict the control output at the next moment. Then, a control strategy integrating Parallel GRU-Transformer feedforward compensation was designed, and the predicted results were introduced into the feedforward compensation control strategy to compensate for the output of the feedback controller. The effectiveness of the proposed method was verified by comparing its prediction accuracy with the State-of-the-Art neural network model and the original control strategy. The experimental results show that the neural network based on Parallel GRU-Transformer has excellent prediction ability. Introducing its prediction results into the feedforward compensation control strategy significantly reduces tracking error and its Moving Average (MA) and Moving Standard Deviation (MSD), improving tracking performance. Meanwhile, under different reference trajectories and controller conditions, the control strategy exhibits a certain level of robustness, demonstrating notable potential for industrial applications.

To address the limitations of convolutional neural networks in deep feature extraction, this study proposes a dynamic deformable convolution with a four-way parallel structure in the encoder. The method adaptively adjusts the convolutional kernel sampling positions through a learnable offset prediction network, thereby enhancing the model's ability to perceive non-uniform fog distribution. Furthermore, to overcome the constraints of existing attention mechanisms in cross-scale feature modeling, a dual-path collaborative attention feature transformation layer is introduced. This layer improves long-range dependency modeling through interactive information transfer between structural perception and semantic enhancement pathways. Building upon these innovations, this study develops the Dynamic Collaborative Attention Enhancement Network (DCAE-Net). The network achieves precise modeling of haze degradation by leveraging an encoderdecoder architecture and a synergistic design of multi-level feature interaction and dynamic perception units. The key contribution lies in integrating dynamic deformable convolution, cross-modal attention mechanisms, and adaptive fusion strategies to establish a hierarchical processing system with spatial perception capabilities. Experimental results demonstrate that the proposed network outperforms existing methods in preserving image details and color fidelity.

Unmanned-aerial-vehicle (UAV) communications, as a critical application scenario in the sixth generation (6G) wireless communication field, has garnered widespread attention. During UAV-to-ground communication, channel data plays a pivotal role. Analyzing channel data enables an understanding of communication environments' diversity and temporal variability, thereby facilitating the construction of more efficient communication systems. This article proposes a novel UAV-to-ground channel prediction method based on multimodal fusion. The method aims to achieve real-time and precise prediction of UAV-to-ground channel data from UAVs in the 3-D airspace by integrating various sources of information, including UAV-captured images, location data of transmitters and receivers, and communication settings. The network uses a fused architecture combining convolutional neural network (CNN) and Transformer architecture to extract and integrate features from diverse information sources. This fusion strategy significantly enhances the accuracy of UAV-to-ground channel prediction. Incorporating image information enables the network better to comprehend the complexity and dynamics of communication environments, thereby assisting in achieving more precise UAV-to-ground channel prediction. Experimental results demonstrate that the proposed method achieves real-time prediction of ground channels across various flight altitudes and communication frequency bands. This provides robust technical support for advancing UAV communication and offers new insights for optimizing and upgrading future wireless communication systems.

The continuous reduction in critical dimensions and the escalating demands for higher throughput are driving motion platforms to operate under increasingly complex conditions, including multi-axis coupling, structural nonlinearities, and time-varying operational scenarios. These complexities make the trade-offs among precision, speed, and robustness increasingly challenging. Traditional Proportional–Integral–Derivative (PID) controllers, which rely on empirical tuning methods, suffer from prolonged trial-and-error cycles and limited transferability, and consequently struggle to maintain optimal performance under these complex working conditions. This paper proposes an adaptive β–Proximal Policy Optimization with Random Network Distillation (β-PPO-RND) parameter optimization within the Prescribed Performance Control (PPC) framework. The adaptive coefficient β is updated based on the temporal change in reward difference, which is clipped and smoothly mapped to a preset range using a hyperbolic tangent function. This mechanism dynamically balances intrinsic and extrinsic rewards—encouraging broader exploration in the early stage and emphasizing performance optimization in the later stage. Experimental validation on a Permanent Magnet Linear Synchronous Motor (PMLSM) platform confirms the effectiveness of the proposed approach. It eliminates the need for manual tuning and enables real-time controller parameter adjustment within the PPC framework, achieving high-precision trajectory tracking and a significant reduction in steady-state error. Experimental results show that the proposed method achieves MAE = 0.135 and RMSE = 0.154, representing approximately 70% reductions compared to the conventional PID controller.

Satellite signal acquisition is a critical component for BeiDou receivers to achieve PNT functionality, with its performance directly determining system availability in complex environments. Traditional time-frequency two-dimensional parallel search methods face two major bottlenecks: computational complexity increases dramatically when processing high-dynamic Doppler shifts, and under weak signal conditions, they struggle to meet high-sensitivity, low-power requirements due to the squaring loss effect and elevated noise floor. To address these challenges, this paper proposes an innovative acquisition architecture. 1) Combining an equivalent frequency compensation cyclic shift search mechanism with sparse Fourier transform (SFT), eliminating traditional two-dimensional traversal, and transforming Doppler search into cyclic shifts to reconstruct frequency search logic. 2) Introducing SFT to efficiently utilize the frequency-domain sparsity of BeiDou signals, computing only significant frequency components to optimize frequency-domain correlation efficiency. 3) Designing an improved differential coherent integration algorithm. For weak signal acquisition, differential coherent technology is applied by constructing a phase difference model between adjacent symbol periods, effectively canceling data bit transition effects and significantly suppressing noise floor elevation, achieving an SNR gain improvement of approximately 3 dB. Compared to conventional methods, under the same hardware conditions, the proposed solution reduces computational complexity by 62 % and improves acquisition sensitivity by 4 dB. For weak signals at -45 dBm, the acquisition success rate reaches 95 %. The novel “Doppler cyclic shift search-SFT” fusion architecture and differential coherent technology provide a new technical approach for efficient, high-sensitivity BeiDou signal acquisition, demonstrating significant theoretical breakthroughs and broad engineering application prospects.

The safe operation of unmanned ground vehicles (UGVs) demands fundamental and essential requirements for continuous and reliable positioning performance. Traditional coupled navigation systems, combining the global navigation satellite system (GNSS) with an inertial navigation system (INS), provide continuous, drift-free position estimation. However, challenges like GNSS signal interference and blockage in complex scenarios can significantly degrade system performance. Moreover, ultra-wideband (UWB) technology, known for its high precision, is increasingly used as a complementary system to the GNSS. To tackle these challenges, this paper proposes a novel tightly coupled INS/UWB/GNSS-RTK integrated positioning system framework, leveraging a variational Bayesian adaptive Kalman filter based on the maximum mixture correntropy criterion. This framework is introduced to provide a high-precision and robust navigation solution. By incorporating the maximum mixture correntropy criterion, the system effectively mitigates interference from anomalous measurements. Simultaneously, variational Bayesian estimation is employed to adaptively adjust noise statistical characteristics, thereby enhancing the robustness and accuracy of the integrated system’s state estimation. Furthermore, sensor measurements are tightly integrated with the inertial measurement unit (IMU), facilitating precise positioning even in the presence of interference from multiple signal sources. A series of real-world and simulation experiments were carried out on a UGV to assess the proposed approach’s performance. Experimental results demonstrate that the approach provides superior accuracy and stability in integrated system state estimation, significantly mitigating position drift error caused by uncertainty-induced disturbances. In the presence of non-Gaussian noise disturbances introduced by anomalous measurements, the proposed approach effectively implements error control, demonstrating substantial advantages in positioning accuracy and robustness.

This paper presents a new algorithm for real-time positioning and orientation of unmanned surface vessels (USVs) based on a single LiDAR. To address the perception challenges in environments where GNSS works not, this algorithm achieves online and efficient identification of point cloud clusters through dynamic area division, and prioritizes the selection of the point cloud clusters that are most valuable for continuous tracking of USV navigation. By adaptively determining the calculation area based on the real-time position information of the USV, the high-precision reference positions are obtained through point cloud cluster identification, and the position and heading angle of the USV are synchronously solved using geometric constraint. Both the performance and real-time capabilities of proposed algorithm are demonstrated through experiment results, providing an effective solution for the state measurement of USVs in GNSS-constrained environments.

Biomarkers are widely used and accepted for the prediction, diagnosis, and post-treatment monitoring of critical diseases, in both clinic and research. However, the rapid determination of biomarker levels in fluid samples remains a significant challenge. Point-of-care testing is an emerging platform that enables instant and accurate biomarker analysis by integrating advanced biosensors with portable signal analysis systems. In this study, we developed a biosensing platform by combining a flexible sensing array, an immuno-microfluidic biochip, and a signal acquisition/processing unit for rapid protein biomarker detection aimed at cancer screening. The platform consists of an immuno-microfluidic biochip for biomarker capture and pressure sensor activation, a flexible sensing array for signal detection, and a signal acquisition/processing unit to analyze and display the results. The microfluidic biosensing platform achieved a limit of detection of 40 pg/ml and a linear operating range from 0.1 to 150 ng/ml (R-2 = 0.98) in blood serum samples, with excellent stability and selectivity. The platform can process up to 10 samples simultaneously within 48 minutes. Clinical testing of cancer samples showed a correlation of similar to 98.7 % between the biosensing platform and commercial equipment, confirming its high accuracy. This biosensing platform offers an advanced approach for biomarker analysis in cancer screening and diagnosis, with potential applications in the detection and monitoring of other diseases.