Trajectory prediction is one of the core functions of autonomous driving. Modeling spatial-aware interactions and temporal motion patterns for observed vehicles are critical for accurate trajectory prediction. Most recent works on trajectory prediction utilize recurrent neural networks (RNNs) to model temporal patterns and usually need convolutional neural networks (CNNs) additionally to capture spatial interactions. Although Transformer, a multi-head attention-based network, has shown its notable ability in many sequence-modeling tasks (e.g., machine translation in natural language processing), it has not been explored much in trajectory prediction. This paper presents a Spatial Interaction-aware Transformer-based model, which uses the multi-head self-attention mechanism to capture both interactions of neighbor vehicles and temporal dependencies of trajectories. This model applies a GRU-based encoder-decoder module to make the prediction. Besides, different from methods considering the spatial interactions only among observed trajectories in both encoding and decoding stages, our model will also consider the potential spatial interactions between future trajectories in decoding. The proposed model was evaluated on the NGSIM dataset. Compared with other baselines, our model exhibited better prediction precision, especially for long-term prediction.

As an important carrier of individual information, the resume is an important data source for studying the spatio-temporal evolutionary characteristics of individual and group behaviors. This study focuses on spatio-temporal information extraction and geoparsing from resumes to provide basic technical support for spatio-temporal research based on resume text. Most current studies on resume text information extraction are oriented toward recruitment work, such as the automated information extraction, classification, and recommendation of resumes. These studies ignore the spatio-temporal information of individual and group behaviors implied in resumes. Therefore, this study takes the public resumes of teachers in key universities in China as the research data, proposes a set of spatio-temporal information extraction solutions for electronic resumes of public figures, and designs a spatial entity geoparsing method, which can effectively extract and spatially locate spatio-temporal information in the resumes. To verify the effectiveness of the proposed method, text information extraction models such as BiLSTM-CRF, BERT-CRF, and BERT-BiLSTM-CRF are selected to conduct comparative experiments, and the spatial entity geoparsing method is verified. The experimental results show that the precision of the selected models on the named entity recognition task is 96.23% and the precision of the designed spatial entity geoparsing method is 97.91%.

Event detection (ED) aims to identify events of interest described in the text. With the current explosive growth of text data on the internet, ED is increasingly practical and has gained many researchers’ attention. The existing works usually design ED as a token-level multi-class classification task. In this setting, given a sentence, ED models’ prediction for each token is relatively independent and thus cannot fully utilize sentence-level information and the association relations between multiple events in this sentence. To handle these situations, this paper proposes a multi-task learning based event detection model, which introduces an event type oriented text classification as an auxiliary task to improve the model’s understanding of sentence-level information. In addition, this model utilizes a Conditional Random Field (CRF) to explore the correlations between various event types and constrain the model’s output space. Experimental comparisons with state-of-the-art baselines on DuEE dataset demonstrate the model’s effectiveness.

China often uses equal length sections as the basic unit for section classification, which may lead to the value of PCI (Pavemant Surface Condition Index) is out of line with the actual, causes problems such as waste of maintenance resources or deterioration in the quality of pavement use, based on the principle of ordered clustering and the concept of loss function in Fisher's algorithm, the Fisher optimal. The method of combination of Fisher optimal partition and traditional ordered clustering analysis is applied to road segmentation. The statistical matrix of disease types is used as the variable index, and the error function of class diameter is defined between the indexes to discriminate the optimal division, and the second-order difference statistics method of ordered clustering applicable to pavements is proposed to determine the optimal classification number, and the classification results are further improved according to the characteristics of cement concrete pavement maintenance in China. Through the comparison of testing and validation analysis of engineering examples, the pavement classification method based on ordered clustering unequal division can more accurately reflect the local changes of cement concrete pavement damage condition and provide a more scientific basis for the decision of cement concrete pavement maintenance scheme.

Population activity drives urban development, and high-spatiotemporal-resolution population distribution provides critical insights for refined urban management and social services. However, mixed population activity patterns and spatial heterogeneity make simultaneous high-temporal- and -spatial-resolution estimation difficult. Therefore, we propose the High-Spatiotemporal-Resolution Population Distribution Estimation Based on the Strong and Weak Perception of Population Activity Patterns (SWPP-HSTPE) method to estimate hourly population distribution at the building scale. During the weak-perception period, we construct a Modified Dual-Environment Feature Fusion model using building features within small-scale grids to estimate stable nighttime populations. During the strong-perception period, we incorporate activity characteristics of weakly perceived activity populations (minors and older people). Then, the Self-Organizing Map algorithm and spatial environment function purity are used to decompose mixed patterns of strongly perceived activity populations (young and middle-aged) and to extract fundamental patterns, combined with building types, for population calculation. Results demonstrated that the SWPP-HSTPE method achieved high-spatiotemporal-resolution population distribution estimation. During the weak-perception period, the estimated population correlated strongly with actual household counts (r = 0.72) and outperformed WorldPop and GHS-POP by 0.157 and 0.133, respectively. During the strong-perception period, the SWPP-HSTPE model achieves a correlation with hourly population estimates that is approximately 4% higher than that of the baseline model, while reducing estimation errors by nearly 2%. By jointly accounting for temporal dynamics and population activity patterns, this study provides valuable data support and methodological insights for fine-grained urban management.