The real-time precise point positioning (RT PPP) technique has attracted increasing attention due to its high-accuracy and real-time performance. However, a considerable initialization time, normally a few hours, is required in order to achieve the proper convergence of the real-valued ambiguities and other estimate parameters. The RT PPP convergence time may be reduced by combining quad-constellation global navigation satellite system (GNSS), or by using RT ionospheric products to constrain the ionosphere delay. But to improve the performance of convergence and achieve the best positioning solutions in the whole data processing, proper and precise variances of the observations and ionospheric constraints are important, since they involve the processing of measurements of different types and with different accuracy. To address this issue, a weighting approach is proposed by a combination of the weight factors searching algorithm and a moving-window average filter. In this approach, the variances of ionospheric constraints are adjusted dynamically according to the principle that the sum of the quadratic forms of weighted residuals is the minimum, and the filter is applied to combine all epoch-by-epoch weight factors within a time window. To evaluate the proposed approach, datasets from 31 Multi-GNSS Experiment (MGEX) stations during the period of DOY (day of year) 023-054 in 2018 are analyzed with different positioning modes and different data processing methods. Experimental results show that the new weighting approach can significantly improve the convergence performance, and that the maximum improvement rate reaches 35.9% in comparison to the traditional method of priori variance in the static dual-frequency positioning mode. In terms of the RMS (Root Mean Square) statistics of positioning errors calculated by the new method after filter convergence, the same accuracy level as that of RT PPP without constraints can be achieved.

Accurately forecasting air pollutants concentration can reduce health risks and provide an important reference for environmental governance. This study proposes a new deep learning model, GLA-Net, which aims to achieve high-precision prediction of the air quality index (AQI) of monitoring stations. Specifically, GLSTM-Block is designed to use the GAT module to capture dynamic spatial interaction of AQI, generate spatial semantic features, and then use LSTM to capture the temporal correlation characteristics of these spatial characteristics. This paper uses an LSTM network outside of GLSTM-Block to capture the original temporal characteristics of the input data. Then, the temporal characteristics of the LSTM output are added to the dynamic spatiotemporal features of the GLSTM-Block to obtain the final spatiotemporal features as the input of the subsequent temporal attention layer. The temporal attention layer uses a multi-head self-attention mechanism to focus on the impact of the spatiotemporal characteristics of historical air quality data on each prediction time step, and performs AQI prediction through a fully connected layer. Analysis based on measured data from Beijing, Shanghai and Shenzhen shows that the GLA-Net model has significant advantages in predicting single-step and multi-step changes in AQI. The study found that although the model has a large absolute error in predicting concentrations in highly polluted areas, it can better grasp the trend of changes. This feature is particularly evident in Beijing (AQI mean 64.289), with root mean square error (RMSE) of 12.716 and index of agreement (IA) of 0.983.

An unscented total Kalman filter (UTKF) estimator with nonlinear dynamic errors-in-variables (DEIV) model is derived based on correlational inference. The proposed UTKF considers all random errors in both system and observation equations and is a Jacobian matrix free alternative to the existing TKF estimators. In particular, this estimator is applied to the inertial navigation system (INS)/ultra-wideband (UWB) integration, in which the marginalised unscented transformation (MUT) as well as the use of generalised Rodrigues parameter (GRP) for attitude updates are embedded into the UTKF to improve the computational efficiency and deal with the dimensional mismatching problems. Furthermore, a theoretical analysis to the effects of DEIV model on total Kalman filter is given. Simulation test has been conducted to compare the performance of UTKF and standard unscented Kalman filter (UKF) in terms of attitude, velocity and position errors. The results demonstrate the feasibility and effectiveness of the proposed estimator.

Effectively dealing with multiple outliers is indispensable for achieving reliable navigation and positioning performance in GNSS. The iterative data snooping (IDS) method is commonly used to identify outliers in a forward stepwise manner. Nevertheless, its effectiveness is often compromised by masking and swamping when multiple outliers are present. Recently, the iterative data refining (IDR) method was proposed as a backward alternative to IDS, which identifies potential inliers from the initial outlying set. However, its performance depends critically on the preset maximum number of suspected outliers (qmax), as this critically affects the quality of the initial outlying set. A large qmax covers all outliers at the expense of estimation accuracy, whereas a small qmax may fail to mitigate the masking and swamping effects. To address this dilemma, a dynamic adjustment strategy for qmax is adopted. Meanwhile, random sample consensus (RANSAC) is employed to evaluate whether the resulting inliers obtained by IDR still contain outliers, which guides the adjustment of qmax. Therefore, this paper proposes a RANSAC-based IDR method that dynamically adjusts qmax in a piecewise increasing manner. Taking GNSS single-point positioning as an example, the proposed method is compared with least squares (LS), IDS, information criterion-based model selection (DIA-AICc), and IDR with fixed qmax under various gross error simulation strategies, using one day of observation data. Results indicate that the average positioning accuracies of the four methods are 97.250 m, 81.075 m, 22.500 m, and 16.224 m. In terms of solution availability, the average results are 100% for IDS, 89.29% for DIA-AICc, and 93.90% for the proposed method, while the average wrong identification rates reach 57.78%, 20.24%, and 12.72%, respectively. Overall, the proposed method consistently achieves the highest correct identification rate and the lowest wrong identification rate among all compared methods, with competitive positioning accuracy and acceptable computational time.

Attention mechanisms have gained significant prominence in the realm of natural language processing. When it comes to relation extraction in the form of a graph, the complex sentence structure and uncertain entity positions pose considerable challenges. To address these limitations, we propose a hierarchical attention graph convolutional neural network (HAGCN) for relation extraction. The HAGCN model reflect the extraction process at multiple levels spanning the graph, sentence, and word levels; In addition, the proposed model incorporates three levels of attention mechanisms which precisely adjust feature vector distribution in the semantic space to prioritize essential vocabulary information and facilitate the extraction of dependency distances between nodes and entities, enhancing the comprehension of intricate sentence structures. The experimental results of TACRED and SemEval indicate that the proposed model performs better than other baseline models.