Ying Xu

Significant wave height (SWH) is a key parameter that characterizes the energy and state of ocean surface waves, playing a vital role in navigation safety, marine engineering design, and studies of air–sea interactions. Spaceborne global navigation satellite system reflectometry (GNSS-R) holds great potential for monitoring ocean dynamic environments, among which the accurate retrieval of SWH is a critical application. In this study, we developed a machine learning–based retrieval framework that integrates multi-system (C/G/R/E), multi-polarization (LHCP/RHCP/H/V), and incidence-angle partitioning strategies using data from the Tianmu-1 (TM-1) constellation. The framework aims to enhance the stability and physical consistency of SWH retrieval. Three gradient boosting models—CatBoost, LightGBM, and XGBoost—were employed, with Bayesian optimization used for adaptive hyperparameter tuning. The experiments were conducted over the Gulf of Mexico, where ERA5 reanalysis data served as the reference truth for model evaluation. Results indicate that the proposed framework effectively captures the nonlinear relationships between delay Doppler map (DDM) features and SWH. All three models achieved optimal performance under moderate incidence angles (10°∼40°). Distinct differences in retrieval accuracy were observed across GNSS systems and polarization modes, with the C and E systems yielding the best results. Specifically, the C-S4 and E-S1 polarization combinations produced the lowest errors (RMSE = 0.48 m). In the fusion experiments, the model combining multiple polarization modes achieved the highest overall accuracy (RMSE = 0.505 m), exhibiting a more compact and symmetric error distribution, thereby demonstrating enhanced robustness and generalization capability. The SHAP-based interpretability analysis further revealed that DDM reflected power, skewness, and spread parameters were the dominant predictors, elucidating the correspondence between model outputs and the underlying ocean surface scattering mechanisms. Overall, this study verifies the feasibility and superiority of the proposed multi-system, multi-polarization, and incidence-angle–aware machine learning framework for spaceborne GNSS-R SWH retrieval, providing a solid methodological and theoretical foundation for global-scale ocean parameter inversion.

Ultra-wideband (UWB) technology has been widely adopted for indoor positioning due to its high temporal resolution. However, the accuracy of UWB-based indoor positioning is fundamentally limited by ranging measurement errors, particularly under non-line-of-sight (NLOS) conditions, where systematic bias and uncertainty are introduced into the measured distances. In this paper, a measurement error mitigation method is proposed to improve UWB ranging reliability in complex indoor environments. The method first identifies NLOS measurements using low-dimensional physical features and a lightweight machine learning classifier. Subsequently, an error compensation strategy is applied to correct biased ranging observations, which are then incorporated into a nonlinear least squares positioning model. Experimental results obtained in typical indoor environments demonstrate that the proposed method significantly reduces ranging errors and improves positioning accuracy compared with conventional approaches. The results indicate that the proposed framework effectively enhances measurement robustness without increasing system complexity.

Reliable positioning and mapping in roadway tunnels are crucial for vehicle-based monitoring and inspection, especially considering the challenging environmental conditions such as rapidly changing illumination, low-texture environments, and repetitive structural elements. While general LiDAR-inertial odometry (LIO) frameworks and loop-closure detection methods are effective in general scenarios, they often suffer from severe drift or incorrect loop constraints under these specific conditions. These challenges are further exacerbated by the inherent uncertainties associated with low-cost sensors. This paper introduces a narrow field-of-view LiDAR-centric RTK-visual-inertial SLAM system enhanced by three key modules: semantic-assisted loop detection and matching, two-stage RTK quality control, and adaptive factor graph optimization (FGO). In the first module, the proposed semantic loop descriptor (SLD) matching is used to determine the potential loop closure locations and then integrates the corresponding constraint as graph nodes. The quality control module addresses RTK outlier rejection during tunnel entry and exit, employing an event-driven stochastic model to characterize the uncertainty between RTK and the other sensors, effectively suppressing RTK-induced errors. FGO module performs optimization by incorporating LIO, RTK, and loop closure factors, employing a keyframe-based strategy to produce globally optimized poses while continuously updating the map. The proposed Tunnel-SLAM was evaluated against state-of-the-art SLAM algorithms in four extended roadway tunnels, ranging in traveling distance approximately from 5 to 10 km. Experimental results demonstrate that the proposed SLAM achieved a final drift of less than 2 m with loop closure, demonstrating significantly reducing the drift, while other existing SLAM frameworks fail catastrophically or have large drift.

High-precision underwater active-sonar navigation system faces an ill-posed positioning problem in request-response localization mode, compounded by prohibitive costs and computational demands associated with in-situ sound speed profile (SSP) measurements. This study presents a novel constant sound speed (CSS)-based acoustic navigation model that is enhanced through sliding-window processing that incorporates Zenith Acoustic Delay (ZAD) compensation and depth constraints. This model eliminates the need for in-situ SSP acquisition while improving positioning accuracy by reducing systematical error influence. The core innovation jointly estimates ZAD parameter to compensate spatiotemporal sound-speed variations within a sliding window and utilizes a B-spline trajectory model to map two-way travel-time (TWTT) observations onto vehicle trajectories. Experimental validation, including simulations and field data from stations at approximately 1700 m and 2200 m depths, demonstrates consistent meter-level 3D positioning accuracy (RMSE: 1.04-1.75 m) across diverse sound speed conditions (in-situ, historical, and proposed empirical CSS). Furthermore, the proposed CSS-based model achieves approximately 100 times faster computation than traditional ray-tracing-based model by avoiding ray-inversion calculations. Crucially, ZAD compensation significantly reduces positioning error induced by sound-speed variations, delivering accuracy comparable to that of in-situ SSP-based models. Notably, depth constraints from pressure sensors (PS) resolve inherent correlations between vertical position and the ZAD parameter, reducing vertical error by 78%. Acoustic positioning residuals are superior to those of ray-tracing-based positioning. This breakthrough eliminates the fundamental requirement for in-situ sound speed measurements, establishing a versatile and efficient solution for next-generation compact, unmanned, and intelligent autonomous underwater vehicle (AUV) navigation. The model enables real-time, low-cost operation in resource-limited scenarios, including deep-sea exploration, emergency response, and swarm applications.

Ultra-Wideband (UWB) positioning is significantly affected by Non-Line-of-Sight (NLOS) errors, which limit its accuracy in complex indoor environments. This paper proposes a lightweight NLOS error recognition method based on two low-dimensional physical features: the power difference between the total received and the first path,power and the ranging value. Three datasets were collected under different occlusion conditions (wooden door, wall, glass, and human body) across multiple environments. Four representative algorithms—KNN, RF, BPNN, and CNN—were evaluated in terms of accuracy and robustness. Results show that traditional machine learning methods (KNN, RF) consistently achieve 97%–99% accuracy across all datasets and demonstrate strong robustness under cross-scenario conditions, while deep learning models (BPNN and CNN) exhibit limited performance with low-dimensional features, achieving accuracies generally below 65%.The study indicates that low-dimensional features are better suited for machine learning models, whereas deep learning requires higher-dimensional and richer inputs to realize its potential. These findings provide practical insights for error modeling and compensation in low-cost UWB positioning systems.

Tropospheric delay is a major error source in high-precision Global Navigation Satellite System (GNSS) positioning. Slant Path Delay (SPD) exhibits azimuthal non-isotropy, and the resulting errors are non-negligible in high-precision GNSS positioning. Existing non-isotropy quantification methods rely on complete SPD grid data generated by Numerical Weather Models. However, constrained by the spatial distribution of satellites in actual GNSS observations, obtaining the average SPD over the full azimuth domain is difficult, limiting the application of these methods in real-time Precise Point Positioning (PPP). This paper proposes a GNSS PPP-based quantification method for tropospheric delay non-isotropy, which directly solves for the non-isotropic value characterizing azimuthal differences using GNSS observation data. The method constructs equations using differences in carrier-phase residuals between satellites at similar elevation angles. Combined with multi-epoch combined adjustment and least squares solution, it overcomes the dependence on data with full azimuthal coverage. Experimental results show the average RMS of the slant delay, computed using the non-isotropy value with VMF3 (Vienna Mapping Function 3), is 9.30 cm. Within the 10°–50° elevation range, it demonstrates stable accuracy performance, with an RMS ranging from 7.75 cm to 9.47 cm. After introducing the non-isotropy value into static PPP, positioning accuracy improved by 12.42%, 4.93%, and 10.57% in the East, North, and Up directions, respectively, and convergence time was reduced by an average of 17.50%. Improvements were particularly significant in areas with complex terrain. This method provides a new approach for modeling tropospheric delay in GNSS positioning, supporting meteorological research and high-precision location-based services.