To address the challenges of abrupt curvature changes, high collision risks, and repeated adjustments in parallel parking within narrow urban passages, this study proposes a three-segment one-step path planning method integrating curve combinations and numerical optimization to enhance vehicle steering performance and dynamic obstacle avoidance capabilities. Firstly, based on the curvature characteristics of arcs and spiral curves, two types of curve combinations, namely CAC (clothoid-arc-clothoid) and CC (clothoid-clothoid), are designed. Secondly, the parking process is reversed, the CAC curve combination is used to design the path of the end section. Then, the constrained optimization model of the CC curve combination is constructed to plan the path of the starting section. The middle section path is designed by introducing the fivetic polynomial curve. Finally, multi-condition simulation tests are carried out in the quasi-feasible area through Matlab. The results show that this method can plan a collision-free and continuously curved feasible path in a narrow scene with a channel width of only 3 meters, and it occupies a relatively narrow channel width. Comparative analysis reveals significant advantages over quintic polynomial and arc-clothoid-straight methods in path smoothness, obstacle avoidance, and space utilization efficiency. Model Predictive Control (MPC) validation confirms precise path tracking with continuous front-wheel steering angle changes. This method provides an efficient and safe solution for narrow-space parallel parking, effectively alleviating road congestion caused by roadside parking operations and advancing urban traffic efficiency.

Intercity road network accessibility is a key indicator of the efficiency of regional flows of economic and social resources. However, existing studies mainly focus on macro-level connectivity, while local service and economic interaction between nodes receive less attention, limiting support for coordinated regional development. To address this issue, a multi-dimensional accessibility evaluation framework for intercity road networks is proposed. First, a node classification method is developed to balance regional development goals and data availability. It identifies the positions and functions of economic centers, transport stations, and tourist attractions within the network. Second, expressways, national highways, and provincial highways are integrated into a unified topological model. A time impedance adjustment mechanism is applied to reflect travel efficiency and cost differences, capturing the dynamic balance of path selection across road types. Based on this structure, three accessibility evaluation models are constructed from three dimensions: connectivity, service, and economy. They quantify overall network connectivity, residents' convenience in accessing stations and attractions, and the strength of economic interaction between nodes. An empirical analysis of the Guangdong-Hong Kong-Macao intercity road network at the end of 2021 shows that the Greater Bay Area performs significantly better in multi-dimensional accessibility than Eastern, Western, and Northern Guangdong. Following principles of evidence-based design, key nodes are identified and high-standard roads are proposed. The results show substantial improvements in both regional and local accessibility. In Eastern, Western, and Northern Guangdong, 99.2% of nodes show improved connectivity, and 83.3% display enhanced economic accessibility. These results validate both the proposed evaluation framework and the optimization scheme. The study provides a systematic tool and decision-making reference for regional road network optimization and resource allocation.

To explore the spatiotemporal characteristics of collision accidents between electric vehicles and vulnerable road users under mixed traffic conditions, this study proposes a method for identifying spatiotemporal hot spots of collisions between electric vehicles and pedestrians/non-motorized vehi-cles. First, based on collision data involving electric vehicles and pedestrians/non-motorized vehicles, the Analytic Hierarchy Process is employed to determine the weights of the influencing fac-tors. Subsequently, a weighted kernel density estimation method reveals the spatial clustering of traf-fic accident locations. Building on this foundation, the Density Peaks Clustering algorithm is utilized as a spatial clustering model for accidents, and a Spatial Temporal-DBSCAN model is constructed to incorporate the temporal dimension, thereby accurately characterizing the spatiotemporal distribution characteristics of collision accidents between electric vehicles and pedes-trians/non-motorized vehicles. Finally, an empirical analysis is conducted using data from a city over a continuous period of 11 months regarding electric vehicle accidents. The research findings indicate that collisions between electric vehicles and pedestrians/non-motorized vehicles exhibit three tem-poral peaks, differing from the traditional bimodal trend observed in traffic accidents, while spatially, accidents display localized clustering characteristics. In terms of spatial hot spot identification, the DPC model outperforms the DBSCAN, OPTICS, and MeanShift algorithms, showing improvements of 42.9%, 74.5%, and 11.1% in silhouette coefficient, Davies-Bouldin Index (DBI), and Calinski-Harabasz Index (CHI), respectively. For spatiotemporal hot spot identification, under comparable DBI conditions, the ST-DBSCAN model achieves silhouette coefficient and CHI values that are 2.25 times and 57.3% higher, respectively, than the best values obtained from ST-OPTICS, ST-DPC, and ST-MeanShift algorithms.

Traffic state and road network structure have a complex interaction relationship, which affects people's travelling efficiency and satisfaction. In this paper, the number of GPS points and speed of road sections to the speed of free flow at night is taken as an indicator of traffic state, and the coupling relationship between traffic state and the degree of road sections, betweenness centrality, closeness centrality and eigenvector centrality is explored by using the coupling coordination degree model. With the coupling relationship as the dependent variable, thirteen independent variables were selected to regress the mechanism of the coupling degree using CatBoost combined with SHAP. The results show that: the number of intersections, green area, and the number of bus stops promote the coupling coordination between traffic status and road network structure, and the road separation type has a significant positive effect on the degree of coupling between traffic state and betweenness centrality, closeness centrality, and degree; for the positively uncoordinated roads, the intensity of public management land and residential land have a negative effect on the coupling coordination between traffic state and closeness centrality. For positively uncoordinated roads, the intensity of public management land and residential land has a negative effect on the coupling of traffic state with closeness centrality, and the intensity of residential land has a negative effect on the coupling of traffic state with eigenvector centrality; for negatively uncoordinated roads, the intensity of residential land has a negative effect on the coupling of traffic state with closeness centrality. The research results can provide new ideas for the planning, optimisation and adjustment of road network structure.

Accurate CO₂ emission measurement and dynamic forecasting are crucial for achieving China's "dual carbon" goals (carbon peak and carbon neutrality). This study integrates the emission factor method, CO₂-CEMS, and machine learning technologies to propose a CO₂ emission measurement and forecasting method based on multi-source data fusion, aimed at providing an efficient and accurate carbon emission monitoring tool for coal-fired power plants. By comparing the CO₂ emission calculation results of different methods, the study evaluates the performance of machine learning algorithms, such as Adaboost, and develops a dynamic forecasting model based on multi-source data. Experimental results show that in Units 1 and 3 of a power plant in Hebei, China, the relative deviations of CO₂ emission calculations between the emission factor method and CO₂-CEMS were 1.63% and -1.27%, respectively, demonstrating good cross-validation between the two methods. Multiple machine learning models (such as XGBoost, LightGBM, etc.) were compared against the AdaBoost model. In addition to conventional evaluation metrics such as R² and MAPE, a novel selection criterion, namely the mean deviation (x_c) obtained by applying trained models to other units, was proposed to assess the generalization capability of machine learning algorithms for further model screening. The results revealed that AdaBoost exhibited superior performance in prediction accuracy and stability, along with higher generalization capability and robustness. The dynamic CO₂ emission forecasting using the optimized Adaboost algorithm achieved R² values greater than 0.99 on both the training and testing sets, with a mean absolute percentage error (MAPE) below 2%, indicating high prediction accuracy, stability, generalization ability, and robustness. The proposed multi-source data fusion method not only effectively overcomes the limitations of traditional methods in dynamic scenarios but also enables precise hourly CO₂ emission forecasting based on real-time data.

Re-ranking methods is a post-processing technique and have demonstrated significant effectiveness in cross-modal retrieval tasks. By mining and processing information between the initial ranking list, they effectively improve retrieval accuracy. Currently, mainstream cross-modal retrieval re-ranking methods re-rank the initial list based on paired datasets. However, they lack flexibility, as they cannot be easily plugged into existing systems without modifying the original framework and retraining, making them difficult to transfer to other frameworks. Moreover, they cannot be applied in unpaired scenarios. While significant progress has been made in cross-modal retrieval tasks with large-scale paired datasets, the issue of requiring substantial resources to label such large datasets in practical scenarios is often overlooked. To address these issues, this paper proposes a Neighbor Information Aggregation-based Unpaired Cross-modal Retrieval Re-ranking Method. The method improves retrieval performance by mining and utilizing neighbor information of samples, pushing incorrect answers away from the query input. It searches for local neighbors in the Euclidean neighborhood and global neighbor expressions through collaborative expression to gather neighbor information, and integrates these two types of neighbor information to generate new features for re-calculating semantic similarity with the retrieval input to complete the re-ranking process. This method is applied as a post-processing technique in several cross-modal retrieval model frameworks and is tested on the MSCOCO dataset, demonstrating the effectiveness of our method and its superiority over other re-ranking methods.

Light field cameras capture synchronized spatial-angular information, offering a new paradigm for 3D visual perception. Depth estimation, as a fundamental task in light field analysis, underpins critical applications including 3D reconstruction and visual odometry. However, occlusion-induced estimation errors remain a persistent challenge. This paper proposes an occlusion-aware depth estimation framework featuring two novel modules: entropy-based occlusion mask pre-computation method and a viewpoint screening-driven depth estimation algorithm. First, we model light field occlusion by analyzing the entropy of micro-images array, constructing a local entropy extremum-based occlusion mask pre-computation approach. This method addresses the limitations of conventional techniques in characterizing occluded regions. Subsequently, viewpoint screening is employed to eliminate interference from occluded perspectives, effectively reducing estimation errors and decreasing the proportion of disparity outliers exceeding 0.03 pixels compared to other methods. The core contribution lies in establishing a theoretical connection between entropy and light field occlusion, enabling a robust occlusion-aware depth estimation framework. Experiments on 4D light field benchmark demonstrate that our method achieves superior performance in mean absolute error (MAE) and 25th error percentile (Q25). Comparative ablation studies confirm the efficacy of the entropy-driven occlusion mask, highlighting the critical role of entropy in advancing light field occlusion.

Scanning transmission electron microscopy (STEM) can realize the characterization and analysis of the microstructure at nano and atomic scales. However, the imaging process will be affected by unpredictable non-uniform noise, which will affect the atomic structure analysis. In the past few years, predictive models based on deep learning networks can reduce the impact of noise through de-noising or data fitting, but these models still meet an overfitting problem. In this paper, a high-noise STEM atomic structure segmentation method based on structural condition priors of materials is designed by introducing structural condition priors of materials into deep neural network models. The structure condition prior of materials is modeled into segmented network attention using the contrast learning method, which includes the self-attention mechanism and cross-attention mechanism. By calculating these two kinds of attention, the segmentation network can not only focus on the key regions in the image but also focus on the control information from the structure coordinate vector mode. The experimental results show that the proposed segmentation network can segment the atomic structure of high-noise STEM images more accurately than other traditional segmentation networks, and accurately predict the secondary structure information that is blocked by noise or the top layer.

Fast Fourier Transform (FFT), as a fundamental algorithm in scientific computing and signal processing, has been widely applied in digital signal processing, image processing, deep learning, and various other fields. With the growing data scale and increasing processing demands, optimizing FFT algorithms on emerging hardware platforms has become increasingly crucial. This research first conducts an in-depth analysis of Ascend NPU architectural characteristics and their implications for FFT optimization. Based on the matrix-computation-based Stockham FFT algorithm, a series of innovative optimization strategies are proposed: (1) develops a heuristic radix selection algorithm to provide effective radix sequence combinations for different input sizes; (2) designs an efficient computation flow for single-iteration FFT that eliminates the need for real-imaginary separation, significantly reducing global memory access overhead; (3) presents an on-chip cache-based data reading optimization strategy, substantially improving data access speed; (4) implements a data layout optimization method for multiple iterations, effectively enhancing overall memory access efficiency. Experimental results on the Huawei Ascend Atlas 800 platform equipped with Ascend 910 AI processor demonstrate that the proposed optimization strategies achieve an average 4.61-fold speedup compared to non-optimized implementations. Through independent performance analysis and validation of each optimization strategy, results show significant individual speedups ranging from 1.42× to 3.52×, providing valuable technical references for implementing efficient FFT algorithms on emerging NPU architectures.

The present study examines the state of research on vehicle-mounted photovoltaic systems, with a particular focus on the optimization of photovoltaic panel installation area. The research explores the potential for enhancing power generation by optimizing the folding mechanism of the panels. However, the study's scope does not extend to the synergistic optimization of power generation and the additional drag energy consumption of the system. The objective of this paper is to enhance the net power of vehicle-mounted photovoltaic systems. The proposed methodology involves optimizing the aerodynamic performance of these systems. This approach is expected to reduce the drag energy consumption of the system relative to the vehicle, thereby enhancing the net power of the system. In the initial phase of the project, the vehicle-mounted photovoltaic (PV) system, in its folded state, was designated as the object. The aerodynamic high transmittance fairing and the tail wing of the PV system were then designed. Subsequently, three design variables were selected for optimization: the front tilt angle of the fairing, the back angle, and the height of the system. These variables were hypothesized to influence the aerodynamic drag through the construction of an orthogonal experimental table and the use of the analysis of the polar deviation. The system drag, front tilt angle, and back angle are all found to have a monotonic influence on the system aerodynamic drag, as evidenced by the analysis of the main effect plot. In the primary effect plot analysis, the three parameters exhibit monotonic effects on the aerodynamic drag of the system. The structural parameters of the fairing shape are as follows: the forward inclination angle α is 70°, the back angle β is 0°, and the height of the system h is 100 mm. Subsequently, the tail angle of attack of the vehicle-mounted photovoltaic system is optimized, and a three times spline interpolation approximation model is constructed on the experimental data to obtain the optimal tail angle of attack of the lift-to-drag ratio of 33.96°. In comparison with the vehicle equipped with the onboard photovoltaic system that is the subject of this study, the Cd value is reduced from 0.619 to 0.343, representing a 44.59% decrease. Furthermore, the aerodynamic drag is reduced from 484.26 N to 375.56 N, indicating a 22.44% decrease. Additionally, the Cl value is reduced from 0.The range of values from 065 to -0.082 indicates a 226.15% decrease, and the direction of the lift undergoes a transformation from upward to downward. This adjustment serves to mitigate the adverse impact of upward aerodynamic lift on the handling and safety performance of the entire vehicle. A comparison of the original vehicle model with the modified version reveals a significant reduction in Cd value from 0.415 to 0.343, marking a 17.35% decrease. Concurrently, aerodynamic resistance undergoes an increase from 372.42 N to 375.56 N, representing a modest 3.14 N rise. This adjustment effectively mitigates the adverse effects of the onboard PV system on the vehicle's aerodynamic performance. Finally, a comparison and analysis of the net power of the on-board PV system before and after optimization is conducted, revealing a net power difference of 7723.62 W at a vehicle speed of 40 m/s.

Friction stir welding technology is of great significance in improving vehicle crash safety. However, when optimizing the friction welding process parameters for vehicle crash safety, the calculation of explicit dynamics model usually consumes a lot of time and resources. In order to improve the efficiency of process parameter optimization, this paper proposes a vehicle collision time series prediction method based on friction stir welding process parameter analysis. In this paper, the mapping relationship between the welding process parameters and the welding strength is summarized, and based on the data of an SUV body, the finite element method is used to construct the vehicle collision explicit dynamics model. The time series prediction surrogate model is trained by the explicit dynamics calculation results, and the collision safety optimization analysis of friction welding vehicle is carried out. The results show that the proposed method shows high reliability in terms of prediction accuracy, saves 50% of the calculation time compared with the traditional explicit dynamics method, and significantly improves the efficiency of process parameter optimization. The friction welding process parameters are optimized based on the proposed algorithm, which further improves the collision safety of the vehicle and provides an effective reference for the vehicle design and parameter optimization.

Ring-opening polymerization of cyclic sulfur compounds is one of the important methods to synthesize sulfur-containing polymer. Traditional anionic polymerization techniques, however, face challenges due to the high reactivity of sulfur-centered anions, which frequently results in chain transfer side reactions that yield broad molar mass distributions and structural variations. This has the effect of undermining the controllability and reproducibility of material properties. To overcome this issue, this study presents a synergistic catalytic system that combines triethylborane (Et₃B), the phosphazene base ^tBuP₁, and a thiol initiator to facilitate efficient and controlled anionic ring-opening polymerization of propylene sulfide. Experimental characterization showed that adding Et₃B completely blocked chain transfer reactions to monomers and prevented the formation of disulfide bonds, as evidenced by ¹H NMR spectroscopy and size exclusion chromatography (SEC), when the reaction occurred at 0 °C. Density functional theory (DFT) calculations further revealed that Et₃B stabilizes the active sulfur-centered anion intermediates through strong B-S coordination, significantly reducing their reactivity and allowing for precise control over polymerization. Building on this methodology, a “one-pot, two-step” strategy that merges anionic polymerization with reversible addition-fragmentation chain transfer (RAFT) polymerization has been established. An alternating copolymer of carbon disulfide and propylene sulfide was initially synthesized to serve as a macro-chain transfer agent, followed by the introduction of styrene to create a well-defined polystyrene-co-poly(carbon disulfide-alt-propylene sulfide) terpolymer. This work presents a facile strategy for the controlled synthesis of high-performance or functional sulfur-rich polymers.

Suspension crystallization coupled centrifugal separation can directly prepare polyester grade coal-to-ethylene glycol. Based on the narrow boiling, azeotropic and heat sensitive system formed by impurities and ethylene glycol in the production process of coal to ethylene glycol, compared with the traditional distillation purification, suspension crystallization has the advantages of low energy consumption, no solvent and environmental friendly, and can achieve the effect of energy saving and emission reduction under the purification system. The optimization of crystallization operation can provide a reference for suspension crystallization coupled centrifugal separation technology. The effects of seed addition, filter cake thickness, stirring intensity, programmed cooling and mother liquor circulation operation process on the purification effect were studied. The results show that the particle size distribution of crystals is affected by the amount of seed addition. The purity of the product can be improved by increasing the amount of seed added, and the distribution coefficient can be reduced by about 15.8%; Because crystal and sweat gas more fully contact, reduced cake thickness enhances sweating with a maximum reduction of 56.8% distribution coefficient; Stirring intensity directly affects mass and heat transfer. The higher the stirring intensity, the higher the purity of the product, but the yield decreased, and the maximum decrease in distribution coefficient and yield was 39.4% and 30.8%, respectively; Programmed cooling directly affects crystallization time. Due to the increase of crystallization time, the program cooling reduces the purity of the product; The recycling of mother liquor led to the increase of water content in the system, which reduced the degree of under cooling, but improved the purity of the product, and the distribution coefficient decreased by 72.0% when it was cycled for 7 times compared with the fresh mother liquor.