The discipline code system of the National Natural Science Foundation of China (NSFC) serves as the foundation for project application and review management, and provides the framework to guide fundamental research and promote disciplinary innovation. In 2025, the discipline of Architecture and Civil Engineering underwent a systematic review and optimization of its research directions and keyword systems. Specifically, the “E0809 Road and Track Engineering” section has expanded its research scope from six to twenty directions, with a corresponding systematic enhancement of the keyword framework. This revision reflects a strategic transition from conventional engineering disciplines toward green and intelligent systems, resilient safety, systems integration, and engineering solutions for extreme environmental conditions. This paper systematically elucidates the overall ideas and concepts behind this optimization of research directions and keyword systems. Furthermore, it provides an analytical interpretation of future development trends and strategic directions for the discipline.

Recently, significant progress has been made in the construction of domestic and foreign smart highways, which focused on the external perception technologies and vehicle infrastructure collaboration systems. However, there were many limitations including definition, purposes and consistency of framework, making it imperative to optimize the concepts of the traditional road construction, and operation and maintenance. To this end, the development of intelligent transportation system in America, Europe and Japan was summarized in this paper. The experience of these countries was analyzed for reference in our country. And then the typical frameworks of smart highway were given. The physical framework and functional system of auto-intelligence smart highway based on embodied intelligence were proposed, whose intelligent ontology was road infrastructure. Also, it could achieve the empowerment mechanism of whole service life for all-time-all-zone-all-event and meet the functional requirements. First, the auto-intelligence smart highway was seen as an embodied intelligence system. The functional system was established by implementation process analysis method and machine system behavioral control theory. Then, the physical framework was designed based on data types, collection methods, network modes, decisions and service, which could support for five application scenarios. Further, the frameworks and systems of multi-source perception, R2X interaction, and intelligent brain were proposed and their synergistic operation methods were established based on genetic algorithm. Based on embodied intelligence, the data attribute requirements for auto-intelligence smart highway were provided, which could meet the functional purposes. The technology sets and their requirements were also developed, which could achieve fundamental requirements. Finally, the application scenarios were designed, which could lay a good foundation for the sustainable development of construction and operation of smart highway in China. The physical framework and functional system of auto-intelligence smart highway based on embodied intelligence are characterized by clear hierarchy, consistent coordination and good compatibility, whose intelligent ontology is road infrastructure. It could optimize the concepts of the traditional road construction, and operation and maintenance, achieving the empowerment mechanism of whole service life for all-time-all-zone-all-event.

Owing to its significant low-carbon advantages and exceptional mechanical properties, polyurethane binders demonstrate excellent application potential in pavement engineering. However, aging induced by multi-environmental factors, including water, heat, light, and oxygen, represents a critical challenge to the long-term durability of polyurethane binder and pavements. Regarding this issue, this paper systematically reviewed the research progress on the aging and anti-aging of polyurethane materials. The aging mechanism, aging test method, aging evaluation methods, and targeted anti-aging research were summarized and commented based on the requirements of road service. The study showed that the aging of polyurethane is essentially a chemical structural evolution process centered around free radical chain reactions. The generation and transformation paths of active groups are influenced by different environments, thereby affecting the molecular chain breakage, cross-linking, and reconfiguration behaviors. In terms of performance evaluation, existing studies included various methods such as physical properties, static mechanical properties, dynamic viscoelastic properties and microstructure characterization. However, there is a lack of a performance evaluation system that is oriented towards engineering requirements (e.g. high-temperature, low-temperature and fatigue). Molecular structure regulation and targeted anti-aging materials enhance the aging resistance of polyurethane, but the long-term effects on road-use polyurethane binders still require further research. Although substantial progress has been made in aging and anti-aging of polyurethane, the complexity of aging behavior of polyurethane binders used in roadway environments still requires further investigation. Future research should focus on analyzing the multi-source coupled aging mechanisms, constructing and standardizing performance evaluation system towards pavement requirements, designing targeted anti-aging strategy based on aging mechanisms, so as to achieve long-term service of road-used polyurethane binders.

In permafrost regions, accelerated permafrost degradation and thaw settlement are induced by prolonged thermal effects from high-intensity solar radiation on black asphalt pavements, leading to distresses including pavement cracking and differential subsidence. To mitigate the thaw-induced settlement, reduction of cyclic heat absorption in subgrades and maintenance of permafrost thermal stability are of strategic significance for preserving transportation infrastructure and ensuring the socioeconomic sustainability of cold regions. Mature cooling technologies like thermal pipes, ventilated ducts, pile-net embankments are widely implemented for subgrade cooling and preservation of permafrost thermal equilibrium, while research on cooling technologies of pavement engineering in the thermal stability protection of permafrost remains limited in cold regions, with long-term durability deficiencies impeding their widespread application and development. This study comprehensively reviewed recent advances in mainstream pavement cooling technologies, which were consisted of porous pavements, heat-reflective coatings, composite coatings, thermal-resistant pavements and phase-change material pavements, along with the laboratory and field evaluation methods for cooling performance. Firstly, the cooling mechanisms, material compositions, and fabrication methods of each technology were introduced. Subsequently, based on the photothermal and other physical properties of cooling materials and their core action mechanism, the cooling effect and pavement performance of various cooling technologies and their key influencing factors were analyzed. The advantages and limitations of indoor and outdoor cooling performance test methods were concluded around the light source characteristics, equipment function modules and test accuracy of photothermal irradiation experiments. Finally, these pavement cooling technologies were summarized according to the application scenarios and adaptability problems in cold regions. Corresponding improvement measures of various cooling technologies and test methods were proposed addressing the existing shortcomings and scientific issues related to the technologies and testing methods. Prospects and future research directions for the development of asphalt pavement cooling technology systems in cold regions were also suggested.

Polyurethane exhibits outstanding advantages such as super-strong adhesion, controllable performance, adaptability to various substrates, and environmental friendliness, demonstrating great potential to overcome the performance limitations of traditional road bonding materials. However, the water erosion resistance of polyurethane mixtures is insufficient, and the mechanical strength degradation rate is high after freeze-thaw cycles, which can easily cause raveling and pothole damage, seriously affecting the durability of polyurethane pavement. To clarify the degradation mechanism of polyurethane mixture performance under freeze-thaw cycles, this study first analyzed the mesoscopic damage characteristics of the adhesive from perspectives such as contact angle. Subsequently, with the aid of atomic force microscopy (AFM), the microscopic deterioration law of the interfacial bonding between polyurethane and aggregate was investigated, using the modulus of the interfacial transition zone as an evaluation index. Finally, combined with the macroscopic physical indices of the polyurethane mixture, the evolution mechanism of pavement performance was elucidated. The research show that freeze-thaw cycles lead to an increase in microscopic voids inside the polyurethane adhesive and their development into micro-cracks, gradually reducing its thermal stability, mechanical properties, and hydrophobicity. The freeze-thaw failure occurring in the interfacial zone between polyurethane and aggregate contributes more to the overall performance degradation of the polyurethane mixture than the self-damage of the polyurethane adhesive. Under freeze-thaw cycles, the pavement performance of the polyurethane mixture exhibits a segmented trend, with significant degradation in the early stage (a reduction of approximately 67.1%) and moderate degradation in the later stage (approximately 14.5%). The strength loss of the polyurethane mixture during the early stage (0-8 cycles) of freeze-thaw is mainly caused by the decline in the hydrophobicity of the polyurethane adhesive. However, after multiple freeze-thaw cycles (8-16 cycles), the contribution of the polyurethane-aggregate interface damage to the performance degradation of the polyurethane mixture gradually increases. The addition of a silane coupling agent can form “molecular bridges” at the interface. By enhancing the interfacial bonding, the freeze-thaw resistance of the polyurethane mixture can be improved by 3-4 times. The research results are not only significant for improving the freeze-thaw resistance of polyurethane mixtures but also lay a foundation for solving the problem of raveling and potholes in polyurethane pavements.

The morphology of aggregates significantly influences the performance of asphalt mixtures. To overcome the limitations of traditional aggregate morphology analysis methods, machine learning algorithms were introduced to achieve rapid identification and prediction of three-dimensional (3D) morphological parameters based on multi-angle, multi-feature two-dimensional (2D) morphological parameters. 3D morphological parameters and 2D morphological parameters from three projection directions for 400 aggregate particles were obtained by Computed Tomography and image processing. Prediction models for 3D parameters were established based on five machine learning algorithms. The predictive performances were evaluated. The contributions of different input parameters were analyzed by Shapley Additive Explanations. The study reveals significant differences in morphological parameter distributions among different 2D projection planes. The maximum difference between 2D and 3D shape parameter distributions reaches 25.0%. It will result in large errors in distribution of aggregate morphology parameters characterized by solely data from a single projection plane. The Extreme Gradient Boosting (XGBoost) algorithm achieves a balance among predictive performance, generalization ability, and interpretability, with determination coefficients (R²) of 0.86 for the training set and 0.83 for the prediction set. Aspect ratio (I_AR) and shape factor (I_SF) are identified as the core parameters for predicting the 3D needle index (contribution: 89.5%), while circularity (I_Cir) dominates the prediction of sphericity (contribution: 92.1%). All 2D morphological parameters make contributions to predicting the 3D angularity index. Increasing the number of projection planes effectively improves model performance. Compared to models using parameters from a single projection plane, models utilizing parameters from three projection planes improve prediction accuracy by at least 20.0%. Parameters from the horizontal projection plane (top view) contributes almost negligibly to the prediction model of 3D needle index, with an R² below 0.10, indicating limited reliability of top-view morphological analysis alone. These results provide an effective approach for rapidly and accurately identifying the 3D morphological characteristics of aggregates during the design phases of asphalt mixtures, thus facilitating the enhancement of performance through optimized aggregate morphology control.

Aiming at the engineering problem that granite aggregate has insufficient adhesion with asphalt, making it difficult to be applied in asphalt surface layers of high-grade highways. Using granite porous asphalt mixture as the matrix and cement-based grouting material as the filler, a granite-based semi-flexible pavement material and its preparation process were proposed, and the grouting rate was used to evaluate the filling degree of the matrix voids by the cement. By means of small-scale pull-out tests, the interfacial bond strength among aggregate, asphalt, and cement was quantitatively analyzed, and combined with macro- and micro-morphological characteristics of the interface fractures, the multi-phase interfacial interaction mechanism was revealed. Moreover, the pavement performance was systematically evaluated through high-temperature rutting, low-temperature bending, and water stability tests. The test results indicate that the grouting saturation ratio of the granite-based material reached 95.27%, and its interfacial bond strength ranged from 0.42 to 0.85 MPa, representing an increase of approximately 70% compared with that of the basalt-based counterpart. This finding demonstrates that granite can serve as a viable alternative to commonly used basalt as the skeleton aggregate for semi-flexible pavement materials. The interface fractures of the granite-based material are composed of cement, asphalt, and aggregate, and the interaction among the three heterogeneous materials forms a multi-phase interface. Physical adhesion between asphalt and aggregate forms the initial interfacial bonding, while the coupling of physical adsorption, chemical bonding, and mechanical synergy among cement, asphalt, and aggregate enhances the interfacial adhesion; moreover, cement hydration gels interpenetrate with asphalt, forming an asphalt-cement interpenetrating network, which strengthens the interfacial bonding. The Marshall stability, dynamic stability, residual stability ratio, and freeze-thaw splitting strength ratio of the granite-based semi-flexible pavement material are 38.8 kN, 78 535 times·mm^-1, 98.3%, and 99.3%, respectively, which are basically the same as those of the commonly used basalt-based material, but its maximum flexural tensile strain decreases by 6.4% compared with that of the basalt-based material. This study provides an innovative solution and technical support for the efficient application of acidic aggregates in semi-flexible pavements.

To address the limited ultraviolet (UV) aging resistance of raw lignin in asphalt, laccase-activated lignin was prepared using a mild and controllable laccase-catalyzed oxidation method. The regulatory effects of laccase activation on the chemical structure and antioxidant activity of lignin were analyzed. The dispersion state and compatibility of laccase-activated lignin in asphalt were evaluated. The influences of laccase-activated lignin on the rheological properties and aging behavior of asphalt were further investigated. On this basis, the cracking resistance of asphalt mixtures under UV aging conditions was assessed. The results show that the laccase-catalyzed treatment preserves the aromatic backbone structure of lignin while promoting the exposure of phenolic hydroxyl groups, thereby enhancing its free-radical scavenging capacity. At the same dosage, laccase-activated lignin modified asphalt exhibits superior high-temperature rutting resistance, fatigue resistance and cracking resistance potential compared with raw lignin modified asphalt. Laccase-activated lignin effectively retards the photo-oxidative hardening and embrittlement of asphalt. Comprehensive evaluation of rheological properties and aging resistance shows that the optimal dosage of laccase-activated lignin is 6%. Laccase-activated lignin modified asphalt mixtures show significantly lower reductions in pre-peak fracture energy and fracture toughness after UV aging than base asphalt mixtures. Their cracking resistance retention is comparable to that of SBS-modified asphalt mixtures, demonstrating good UV aging resistance and engineering application potential. The results confirm that laccase-activated lignin is a promising biomass-based anti-UV aging modifier for asphalt, providing theoretical support and technical guidance for the development of renewable high-performance road materials.

To improve the self-healing capacity of asphalt pavement under low-temperature conditions and inhibit crack propagation induced by thermal stress, this study selected n-tetradecane phase-change material and an asphalt rejuvenator as the composite core material. Highly methylated melamine-formaldehyde resin (HMMM) was used as the shell material. Microcapsules integrating temperature-regulating and self-healing functions (TR-SI) were synthesized via in situ polymerization. The effects of core-material ratio, emulsifier type and concentration, shear rate, core-to-shell ratio, reaction pH, stirring rate, and reaction temperature on the morphology and encapsulation efficiency of the microcapsules were investigated. The optimal preparation process parameters were determined using response surface methodology, and the properties of the microcapsules were charaterized by scanning electron microscopy (SEM) and nanoindentation, and the techniques. The results showed that the optimal process parameters were as follows: An n-tetradecane to rejuvenator mass ratio of 2∶1, emulsifier D as the emulsifier, an emulsification shear rate of 3 000 r·min^-1, an emulsifier concentration of 5.191%, a core-shell ratio of 1.825∶1, a reaction pH of 5.12, a stirring rate of 600 r·min^-1, and a reaction temperature of 65 ℃. Under these conditions, the microcapsules exhibited a regular spherical shape. The outer surface had a rough outer and a dense, smooth inner shell. The shell thickness ranged from 1.1 to 1.4 μm. The average particle size was 48.24 μm. The coefficient of variation of particle size was 0.195, indicating a uniform size distribution. The average core content reached 80.79%. The nanoindentation hardness of the microcapsules was 0.25 GPa, the Young's modulus was 2.45 GPa, and the mass loss rate at 200 ℃ was only 14.8%, indicating good mechanical and thermal stability. The latent heat of phase change during solidification was 91.1 J·g^-1, and the thermal conductivity was 0.171 3 W·(m·K)^-1. After 200 phase-change cycles, the microcapsule retained more than 90% of their thermal buffering capacity. The formation of the microcapsules involved the adsorption of HMMM prepolymers at the interface of the emulsified droplets, followed by condensation and crosslinking reactions. This process formed a dense shell structure and achieved efficient encapsulation of the core material. The temperature-regulating and self-healing integrated microcapsules prepared in this study provide a new material solution for improving the durability of asphalt pavements in cold regions.

To effectively mitigate temperature-induced distresses in asphalt pavements and further enhance their service performance and lifespan, fundamental temperature-regulating materials with high efficiency potential were selected. The influence of the composition and ratio of the composite regulating-temperature agent on the cooling efficacy and thermophysical properties of modified asphalt was investigated. The crystal structure, thermophysical characteristics, and polarization performance of the composite agent under different processing parameters were analyzed in detail. The optimal material ratio and preparation process for the composite regulating-temperature agent were determined, culminating in the synthesis of a durable composite agent possessing both a porous-hollow structure and spontaneous polarization effect. The long-term regulating-temperature effect of the modified asphalt was comparatively evaluated, the cooling/insulating efficacy of the bidirectional self-regulating-temperature-modified asphalt mixture was quantified, and the synergistic enhancement mechanism of bidirectional self-regulating-temperature modifiers was revealed. The results indicate that tourmaline anion powder T, hollow material F, and porous material D are selected as the basic functional materials for synthesizing the self-regulating temperature asphalt modifier, the optimal mass ratio of T, F, and D is 5∶3∶2. At a 20% dosage, the thermal conductivity, thermal diffusion coefficient, and specific heat capacity of the TFD self-regulating-temperature-modified asphalt are 0.155 5 W·m^-1·K^-1, 0.079 5 mm²·s^-1, and 1.701 J·g^-1·K^-1, respectively. When the dosage of TFD composite temperature-regulating agent is less than 20%, the optimal preparation process is high-energy ball milling at '200 r·min^-1-2 h’, while mechanical stirring is recommended for higher dosages. After regeneration and aging treatment, the discrete coefficients of the self-regulating-temperature-modified asphalt are below 3%, demonstrating significant long-term cooling effectiveness. At a 30% TFD composite agent dosage, the cooling and warming values of the modified asphalt mixture at a depth of 5 cm can reach 8.5 ℃ and 3.6 ℃, respectively. The heat resistance and energy release effects of the regulating-temperature agent synergistically enhance the regulating-temperature efficiency of self-regulating-temperature-modified asphalt. This research lays a solid foundation for extending the service life of asphalt pavements and improving the level of intelligence.

To alleviate the summer high temperature damage of asphalt pavement and urban heat island effect, this study developed a phase-change heat-reflective fog seal with self-regulating temperature function using W-VO₂ as the functional material. Firstly, W-VO₂ was optimized and characterized, and the influence of its content on the storage stability of emulsified asphalt and the road performance of fog seal, such as bonding, skid resistance and impermeability was analyzed. Then, the indoor and outdoor tests were conducted to verify the self-regulating temperature performance of the fog seal. Furthermore, the internal relationship between optical properties and thermal physical parameters (thermal conductivity, thermal diffusivity, specific heat capacity) was established to reveal its self-temperature regulation mechanism. Finally, the correlation model between self-temperature regulation and rutting resistance (rutting depth and dynamic stability) is constructed to clarify its mechanism for improving the rutting resistance of asphalt pavement. The results indicate that the incorporation of W-VO₂ significantly enhances the light response ability of the fog seal. Before phase change (<31 ℃), the reflectivity of each band is lower than that of the traditional fog seal (SFS), which is beneficial to increase the pavement temperature in winter. After phase change (≥31 ℃), the reflectivity is comprehensively higher than SFS, which can achieve summer cooling. In addition, W-VO₂ forms an efficient thermal conduction network in the asphalt matrix, improves the thermal physical parameters of the system, and enhances the thermal regulation ability, and always maintains a dark black appearance, effectively avoiding the glare problems. Among them, the fog seal with 8% W-VO₂ (SFS-W8) has the best performance: In summer, the road surface temperature can be reduced by 7.6 ℃, and the duration of high temperature (>60 ℃) is shortened by 3.5 h; in winter, the road surface temperature can be increased by 5.3 ℃, and it exhibits good environmental adaptability and thermal response stability. According to the model prediction, the fog seal can effectively improve the dynamic stability of asphalt pavement and reduce the rutting depth. This study provides an important reference for solving the problems of “over-cooling” and glare of traditional heat reflective pavement in winter, while endowing the pavement with self-regulating temperature and preventive maintenance functions, demonstrating significant engineering application value.

Inspired by the “light-screening” mechanism of botanical leaves, a bioinspired multilayer optical coating for asphalt pavements has been developed to address the limited efficiency of monofunctional coatings and the insufficient synergistic coupling between optical layers in conventional composites. The system is architected with a dual-functional upper layer optimized for solar reflection and shielding, and a specialized base layer for thermal insulation. Initially, using reflectance, transmittance, and thermal conductivity as key optimization metrics, the optimal compositions and application rates for lignin fiber, cesium tungsten bronze, and SiO₂ aerogel monofunctional coatings were established. On this basis, three multilayer configurations (M1-M3) were engineered by manipulating the spatial sequence of the optical layers, followed by a systematic comparative analysis of their zonal optical properties, visual comfort, and cooling efficacy. Results indicate that optimal reflection, shielding, and insulation are achieved at dosages of 8%, 8%, and 3%-with coating loads of 0.4, 0.4, and 0.2 kg·m^-2 respectively. While the reflective profile is predominantly dictated by the top-layer functional materials and varies with layer arrangement, transmittance remains relatively insensitive to internal spatial changes. Peak optical performance is realized through the reflection-shielding synergistic effect. The cooling capacity follows the hierarchy: reflection-shielding-insulation (M1) > shielding-reflection-insulation (M2) > blended-insulation (M3). Notably, the M1 structure exhibited a near-infrared reflectance of 55.2%, a transmittance of 3.0%, and a lightness value of 61.36. This configuration achieved a maximum field cooling effect of 11.6 ℃, while demonstrating superior integrated performance, including a BPN of 54, a bond strength of 0.28 MPa, and an abrasion life of approximately 210 000 cycles. These findings validate that spatial regulation of functional optical layers is a critical determinant in the synergistic enhancement of optical and thermal properties, providing a prescriptive framework for the design of advanced multilayer cooling coatings for urban infrastructure.

To address traffic safety hazards caused by winter road icing, this study aims to achieve active anti-icing performance and reduction of ice-pavement adhesion of pavement surfaces. Sodium acetate (SA) was used as the salt-storage component, while palygorskite (PAL) served as the carrier to prepare a slow-release salt-storage core (SA-PAL). The core was subsequently encapsulated with a polymer membrane (EC-MA) to fabricate a core-shell slow-release salt-storage pavement material (EMP). Electrical conductivity was used to evaluate the amount of released salt, and the optimal formulation was determined based on the slow-release performance and salt-loading capacity. The freezing point and frozen bond strength of asphalt mixtures containing EMP were further examined, and the anti-icing service life was further predicted. Results show that when the KH570 dosage was 0.75, SA-PAL exhibited the best combination of salt-loading capacity and slow-release performance. The EC-MA film had a glass transition temperature of 7.1 ℃ and a mass loss of only 3.99% at 200 ℃, meeting both the fracture requirements under low-temperature loads and the thermal stability required for asphalt construction. By controlling compaction and temperature conditions, the mechanism of slow release-salt precipitation-ice inhibition in low-freezing-point asphalt mixtures was revealed. When the mineral filler replacement ratio was 50%, the freezing point of the low-temperature compacted group decreased to -2.37 ℃, and the frozen bond strength was reduced by approximately 44%. Based on the calculated salt release rates of a 1 m² asphalt pavement under winter and non-winter conditions, the low-freezing-point asphalt pavement was estimated to maintain an effective anti-icing function for about three years under typical southern climatic conditions.

Functional composite pavements with drainage capability achieve rapid drainage and enhanced resistance to base erosion while maintaining the high load-bearing capacity of concrete pavements, thereby improving overall durability. To address the complexity of the interlayer geometry and the difficulty of its characterization and modeling, the interlayer structure was reconstructed using computed tomography (CT). A multilevel geometric parameter system was established at two-dimensional cross-sectional and three-dimensional scales. The complex interlayer geometry was discretized by introducing the concept of peak units, and its spatial distribution characteristics were quantified using Voronoi theory. The results show that the interlayer structure can be reasonably represented as a collection of peak units. The peak-area fraction mainly ranges from 38% to 48%, while peak height, volume, and footprint area exhibit pronounced long-tailed distributions. The peak-unit aspect ratio (1.0-3.5), anisotropy (0.45), and center spacing (19-23 mm) indicate that the interlayer structure is locally irregular but constrained by the aggregate skeleton, forming a statistically quasi-ordered structure. Correlation analyses show that the interlayer structural characteristics can be classified into three parameter groups: geometric scale, morphological characteristics, and spatial distribution. Based on these findings, two equivalent reconstruction methods using constrained parameter combinations were developed. The parameter combination using center spacing and peak-area fraction as primary constraints and peak height and morphological parameters as auxiliary constraints exhibited higher overall statistical consistency than the geometry-dominated combination, with several indices exceeding 85%. Interlayer mechanical response tests and finite element simulations indicate that interlayer shear resistance is directly related to the geometry of the cement-asphalt interaction layer. The finite element models established using the equivalently reconstructed structures captured the non-uniform load-transfer behavior at the composite interface. The proposed equivalent reconstruction method balances geometric fidelity and reproducibility and provides quantifiable and reconstructable geometric input for interlayer mechanical analysis and structural design of composite pavements.

To investigate the damage evolution of pervious pavement concrete under transient hydrodynamic pressure induced by vehicle wheel loads in rainy environments, a self-developed dynamic water dissolution apparatus was used to simulate the dynamic water erosion of pavement surfaces. Pore water pressure sensors were employed to measure the variation patterns of hydrodynamic pressure intensity under different loads and speeds. Next, the degradation patterns of physical properties (mass loss, permeability coefficient, and ultrasonic damage) and mechanical properties (compressive strength, splitting tensile strength) of pervious concrete under various durations of dynamic water dissolution were studied. Furthermore, X-ray diffraction (XRD) and scanning electron microscopy (SEM) were used to reveal the effects of dynamic water dissolution on micromorphology and phase composition, clarifying the damage deterioration mechanism of pervious concrete under dynamic water erosion. The results show that the transient hydrodynamic pressure induced by vehicle wheels is closely related to wheel load and vehicle speed. A quantitative model relating pressure to speed and wheel load was established through fitting. With increasing hydrodynamic pressure, the physical and mechanical properties of concrete deteriorated. After 16 hours of dynamic water dissolution, the maximum mass loss rate of concrete specimens reached 4.84%, the maximum ultrasonic damage rate reached 16.31%, the permeability coefficient increased by up to 36.98%, and the maximum strength loss rates for compressive strength and splitting tensile strength were 14.13% and 19.86%, respectively. Dynamic water action caused surface erosion of the concrete specimens and enhanced internal pore connectivity. The calcium silicate hydrate (C—S—H) gel and calcium hydroxide (CH) crystals within the specimen are dissolved by hydrodynamic leaching, leading to structural loosening and strength reduction of the material. The deterioration of pervious concrete in a hydrodynamic environment is governed by the coupled effect of physical scouring and chemical dissolution. The former causes surface material spalling and pore connectivity, while the latter induces the dissolution of hydration products. Together, they lead to structural loosening and degradation of mechanical properties. This study simulates the service performance of pavement under actual rainy conditions, and the findings provide a theoretical basis for the design of pervious pavement concrete.

Road markings are key information carriers for lane recognition and localization in autonomous driving. However, vision-based detection algorithms often fail under poor lighting, rain, fog, or occlusion, reducing perception accuracy. To enhance reliability under complex conditions, this study proposed an optimized metal reflective road stud for millimeter-wave radar perception. The electromagnetic scattering mechanism was analyzed based on material and geometry, and experiments were conducted for verification. Results showed that: ① Traditional markings have low radar cross-section (RCS) at 77 GHz due to poor conductivity and rough surfaces, while aluminum material improves reflectivity and resists roughness variation. ② The triangular reflective structure greatly enhances echo strength. Vertical plate length (a) and height (h) are the most influential parameters; the proposed design achieves -9 dB RCS at 0°-10° incidence, higher than conventional markings (-19 dB). ③ The Multilayer Perceptron Model built on simulation data predicts RCS from structural parameters with a relative error of 3.5%. ④ Measured and simulated results show consistent trends, confirming the model's reliability and design feasibility. The proposed metal reflective road stud provides a robust solution for radar-based lane perception and infrastructure sensing in autonomous driving.

Regional segmentation is an effective approach for precise detection of asphalt pavement gradation segregation. To address the insufficient long-short term feature interaction and the disrupted local correlations in visual Mamba architectures during gradation segregation detection, this study proposes a segmentation framework based on VM-UNet that integrates feature clustering and a deformable visual state space module. First, a contextual clustering selective state space module was constructed. This module achieved adaptive aggregation of local features through a feature clustering algorithm. Simultaneously, by designing an adaptive scanning strategy, it established a multi-scale feature interaction mechanism that connects global contextual features across different self-organizing windows. Second, a group normalization-enhanced linear attention unit was designed. This unit fused LePE and RoPE encoding mechanisms synergistically, forming a hierarchical receptive field expansion strategy. Third, a multi-channel feature fusion network was organized. It employed dynamic gating and a multi-branch residual structure to achieve complementary fusion of cross-level features. Finally, considering the morphological characteristics of segregation regions, a MorphoFocalDice loss function incorporating morphological constraints was proposed. This loss function was combined with watershed post-processing for instance-level segmentation. Experimental results demonstrated that compared to the baseline model, integrating the contextual clustering adaptive selection strategy enabled adaptive fusion of global and local features, improving mIoU and Acc by 7.8% and 5.6% respectively. Introducing the group normalized linear attention module expanded the receptive field and alleviated the semantic information loss caused by pooling, improving mIoU and Acc by 1.0% and 2.3% respectively. Introducing the efficient feature fusion network using the dynamic multi-channel visual transformer improved mIoU and Acc by 2.4% and 2.8% respectively. The MorphoFocalDice loss function contributed to improving semantic segmentation accuracy, increasing the model's mIoU and Acc by 6.6% and 4.9% respectively. Compared to mainstream models like UNet, DeepLabV3+, and Swin U-Net, our model achieved 8.1% to 18.2% higher mIoU and Acc. While maintaining the advantage of linear computational complexity, it significantly enhanced multi-scale feature representation. The segmentation accuracy and efficiency were superior to comparative methods, demonstrating promising engineering application prospects.

This paper addressed the detection of underground cavity hazards and the classification of surface risk events in urban roads. A multi-risk evaluation method was proposed based on distributed acoustic sensing (DAS) and acoustic emission (AE) signal features. First, a phase cross-correlation-based reliability index β was introduced to evaluate the quality of DAS channel data and initially identify potential underground cavities. Meanwhile, considering the similarity between DAS and AE in acoustic signal processing, 31 common AE-related time-domain, frequency-domain, and composite features were extracted from DAS signals. Pearson correlation analysis was applied to remove redundant features and optimize feature dimensions. Furthermore, an anomaly detection method was developed using kernel density estimation and two-dimensional Gaussian distribution fitting based on a single feature. A confidence interval was determined, and channels outside this interval were identified as anomalies. Principal component analysis (PCA) and DBSCAN clustering were then used to further locate anomalous channels. The union of the results from both methods was taken to identify underground abnormal regions. Finally, statistical analysis of the β-value distribution was conducted. The mean and variance were used to quantify the characteristics of surface risk events for classification. PCA and K-means clustering were further applied to classify different events. The clustering results were used to validate the effectiveness of β in distinguishing risk events. Experimental results show that the proposed method effectively detects underground cavity anomalies and classifies surface risk events, providing a new technical approach for intelligent monitoring and early warning of urban road safety.

Throughout the road life cycle, underground hidden void disease may cause secondary disasters if it is not detected and recognized in time. Ground-penetrating radar (GPR) is a key tool for accurately identifying road defects, but challenges persist, including limited datasets, excessively complex models, and insufficient accuracy. To enable intelligent void detection, this study constructed a high-quality standard dataset of B-Scan images based on an idealized electromagnetic wave propagation model and field acquisition. This study implemented quantitative analysis of void defect images in a simulated environment. To overcome the limitations of current methods, this paper proposes a detection approach based on the YOLO11 algorithm called YOLO11-L. Firstly, this paper optimized the C3k2 module using Partial Convolution (PConv), which applies spatial feature extraction via conventional convolution only to a part of the input channels. This strategy improved computational efficiency while reducing computation and memory access requirements. Second, this paper replaced the Conv module with Adaptive Down-sampling (ADown), which dynamically adjusts receptive fields through learnable parameter to capture multi-scale features. This module effectively reduces feature map dimensions while preserving critical information. Finally, this paper refined the C2PSA module by integrating Efficient Multi-Scale Attention (EMA), which employs parallel substructures to reduce sequential processing depth. This enhancement improves accuracy while reducing parameter counts and boosting model efficiency. The YOLO11-L model achieves a remarkable 97.9% in mAP@0.5, with mAP@0.5:0.95 elevated to 74.9% and mF₁ improved to 97%. It has only 78.5% of parameter of the YOLO11 model. The FLOPs represent just 79.3% of the original model, while inference speed doubles. Additionally, the weight file size is reduced by 1.18 MB from the original. Our approach offers significant advantages: small model size, low complexity, high detection accuracy, strong learning ability, and generalization performance. It is suitable for intelligent void detection in road structures.

The deformation and failure mechanisms of geotextile-encased stone columns (GESC) remain insufficiently understood in existing research due to the complex interactions between granular particles and geotextiles. This study established a GESC model based on a discrete-continuum coupled numerical method. The validity of the numerical approach was verified through comparison with model test results, and the deformation and failure characteristics of the system under triaxial test conditions were investigated. The study explored how macroscopic deformations and failure characteristics relate to changes in the mesoscopic structure. The effects of variations in the geotextile's modulus and tensile strength on the GESC's mechanical properties were analyzed, shedding light on the underlying mechanical mechanisms. The results revealed that the confining pressure solely influences the GESC at first. As loading continues, the constraining effect of the geotextile gradually becomes significant until the geotextile assumes the primary confining role, at which point the stone specimen reaches its critical state. At the ultimate tensile strength, the geotextile ruptures, causing a sharp decrease in the GESC's load-bearing capacity. The study also examined the evolution of coordination numbers, porosity, and force chain networks, identifying distinct phases of interaction between the stone and the geotextile. These stages include initial compaction, a transition stage, critical shear dilation, and post-peak failure. The tensile strength of the geotextile exhibits a linear positive correlation with the peak strength of the GESC, while the peak strength of the GESC increases synchronously with rising confining pressure. The bulging deformation of the GESC aggravates with enhanced geotextile tensile strength. Increased geotextile modulus significantly improves the GESC's modulus during the critical dilatancy stage, demonstrating a linear positive correlation, while the modulus parameters of the GESC show insensitivity to confining pressure. Furthermore, the bulging deformation intensifies with decreasing geotextile modulus.

Various U-bar reinforcement layouts are currently used at wet joints of precast concrete bridge decks in engineering practice. To investigate the influence of different U-bar configurations on the mechanical behavior at the joint, this study designed three full-scale specimens with strictly centered interlaced, closely interlaced, and closely welded U-bar layouts for axial tensile testing. The experimental findings were further supported by finite element simulations using ABAQUS. Results show that initial cracking occurred at the joint interface, followed by sequential cracking in the precast panel and the cast-in-place section. The load-displacement response exhibited a linear-elastic stage, an elasto-plastic stage, and a yielding stage. The U-bar layout had minimal influence on the cracking load and ultimate tensile capacity. Compared to the closely interlaced configuration, the strictly centered layout demonstrated slightly better cracking resistance and load capacity. While welding the U-bars slightly improved the cracking resistance, the effect was limited. Finite element analysis indicated that, provided the spacing and overlap length of U-bars on each side are reasonable, varying the interlacing offset alone has negligible effect on the joint's mechanical behavior and load-bearing capacity. The calculated ultimate capacity based on a strut-and-tie model agreed well with both experimental and numerical results, suggesting that a unified strut-and-tie model can be used for different U-bar configurations. Overall, to improve construction efficiency, strict centering or welding of U-bars is not necessary in engineering applications.

Given the significant variation in collapse risks from ship collisions for long-span cross-sea bridges across different service periods and structural span types, and considering that current methods for determining representative ship types rely exclusively on static waterway vessel data, there is a critical need to develop a scientifically robust and rational approach for defining dynamic anti-collision ship types. A research method combining theoretical analysis, numerical simulation, and engineering case studies was adopted to identify anti-collision ship types relevant to bridge collapse, taking into account structural characteristics of bridges and stiffness degradation of piers. Key achievements include: An energy dissipation ratio model for impacted piers and a ship speed threshold formula were derived, and a quantitative relationship between the ultimate lateral resistance of piers and critical ship parameters (tonnage and speed) for bridge collapse was established. The theoretical framework was validated through a case study on non-navigable spans of a coastal bridge, where computational results show less than 10% error compared to numerical simulations. Furthermore, navigation speed thresholds for ships of various tonnages and the lateral resistance of piers were calculated for the case study bridge under different pier stiffness degradation levels. The resulting collapse-prevention design vessel for this bridge section had a tonnage of 7 000 t and a speed threshold of 2.21 m·s^-1. Further analysis revealed that as pier stiffness degraded, the tonnage of the design vessel at the same speed threshold decreased significantly, with the rate of reduction accelerating gradually. The proposed computational method enables rapid determination of anti-collision ship types, thereby providing references for collapse risk assessment and ship collision prevention design in other existing bridge structures.

The unsteady evolution of the flow pattern around the engineering structure, characterized by vortex shedding, drifting, and other key flow features, determines the spatio-temporal distribution pattern of aerodynamic forces on the structure's surface and the vortex-induced vibration (VIV) response. The vortex-induced aerodynamic forces on the engineering structure's surface are completely expressed into two superposed spatio-temporal modes of aerodynamic forces: a travelling wave mode associated with vortex motion and possessing compulsive properties, and a hysteresis mode exhibiting overall hysteresis relative to the structure's motion dominated by coherent flow structures that are either stationary or moving slowly and possessing self-excited properties. Furthermore, a deconstruction framework for vortex-induced aerodynamic forces, namely the Travelling Wave Mode Decomposition (TWMD) algorithm, is proposed. A typical case of streamlined closed-box girder is used for validation, and the evolution characteristics of the spatio-temporal distribution pattern of aerodynamic forces throughout the entire process of torsional VIV lock-in range are studied by the above means, providing an in-depth explanation of the mechanism of torsional VIV. The research shows that the proposed TWMD algorithm can decompose vortex-induced aerodynamic forces into a travelling wave component dominated by vortex drifting and a hysteresis component unrelated to vortex drifting. During the VIV, the upper surface is dominated by the pressure travelling wave component, while the lower surface is dominated by the hysteresis component. Moreover, the travelling wave component of the global vortex-induced moment is much smaller than the hysteresis component, further confirming that VIV is primarily self-excited and possesses both self-excited and compulsive properties. The phase difference between the upper and lower surface pressure travelling wave components and the global vortex-excited moment decreases monotonically along the downstream, corresponding to 2.5 and 1.5 torsional vibration periods, respectively, indicating significant vortex drift. Moreover, the phase differences in the above regions are almost parallel to each other in space, that is, the same dimensionless coordinate (X/B), and the difference between the phase differences in the above regions is about 360°. The vortex intensity and velocity ratio (the ratio of vortex drift velocity to incoming wind speed) in the upper and lower surface regions are positively and inversely correlated with the VIV amplitude, and the extreme value is obtained at the extreme point of the amplitude. It indicates that there is a significant vortex synergy effect between the upper and lower surface vortices. The research proposes a general spatial-temporal mode framework for vortex-induced aerodynamic forces in engineering structures, quantifying the self-excited effect and the associated vortex interaction effect of vortex-induced aerodynamic forces, laying a foundation for the analysis of the physical mechanism of VIV in engineering structures and the construction of mathematical models for vortex-induced forces.

The bending deformation curvature of main girder is an important performance indicator closely related to bridge damage. Previous studies have used changes in modal curvature for damage identification, but traditional methods require complex modal shape calculations and have limitations such as low accuracy and a high demand for sensors. Therefore, this paper proposes a bridge structural damage identification method based on the dynamic displacement curvature energy difference, aiming to simplify the calculation process of the deformation curvature indicator, reduce the interference of non-sensitive frequency bands on the identification, and improve the accuracy and sensitivity of damage identification. First, based on Euler-Bernoulli beam mechanics analysis and frequency response function theory, the analytical relationship between the displacement response at different frequency ranges and the dominant modal shape is derived. The study shows that for frequency components near the natural frequencies, the displacement responses at all measurement points have the same shape as the modal shape, providing a theoretical basis for the dynamic displacement curvature energy difference indicator. Next, combining statistical analysis of dynamic responses and wavelet packet transform technology, a “two-step” damage identification process is proposed: constructing an initial damage indicator based on the displacement response in the frequency band near the natural frequency and selecting the sensitive frequency bands; further, fusing the dynamic displacement curvature energy difference of the sensitive frequency bands to construct a comprehensive damage indicator for accurate localization of damage. Finally, through damage numerical simulations on continuous beams and progressive damage test analysis of the Swiss Z24 bridge, the accuracy, engineering feasibility, and robustness of the method are verified. The results show that the method can accurately identify single or multiple position combinations of damage and has excellent recognition ability under sparse measurement points and noisy environments, demonstrating promising application prospects.

Road work zones, characterized by temporary lane closures, induce traffic congestion and frequent lane-changing behavior, which may elevate the risk of traffic crashes and serious casualties. They have become the road entities with special concerns in the field of traffic safety research. To systematically reveal the characteristics of driving behavior and traffic safety and their correlation, and to explore the technologies and strategies for safety improvement, the extant literature on driving behavior, crash frequency, risk assessment, crash severity and techniques for safety management and control in work zones is reviewed. This study also discusses the limitations of existing studies and provides some directions for future research. The findings show that: the type and configuration of work zones and the layout of traffic signs and markings, together with the attributes of drivers, road, traffic and environment, significantly shape the driving behaviors like lane-changing decision and speed selection, which then affect real-time crash risk, crash frequency, and crash severity in work zones. In addition to traditional devices such as warning signs, static speed limit signs, traffic cones, emerging technologies including on-board warning systems, variable speed limits, and merge control methods have been proposed for the active safety management of work zones. Nonetheless, further research is needed on the in-depth characterization of driving behaviors under high-risk scenarios, the development of high-precision crash risk prediction models supported by multi-source data fusion, and the design of safety management and takeover strategies in connected and automated vehicle environments.

The relevant studies on dedicated lanes for autonomous driving in recent years were analyzed. According to the overall framework of “deployment-management-application”, the research on lane separation strategies, deployment conditions, lane management and vehicle control, and simulation platform development of dedicated lanes for autonomous driving was systematically reviewed. Based on the review of research progress in each aspect, the impacts of deploying dedicated lanes for autonomous driving were analyzed. The management strategies for dedicated lanes and the vehicle control methods for dedicated lane operations were classified and summarized. The advantages and disadvantages of autonomous driving simulators, as well as the development status of simulation platforms for dedicated lanes, were also summarized. The results show that existing studies across multiple research dimensions generally rely on idealized assumptions, lack sufficient representation of complex mixed traffic conditions and behavioral heterogeneity, and give limited consideration to realistic factors such as communication delays and environmental disturbances. Research on lane separation strategies mainly focuses on the impacts of different designs on traffic operational stability, while in-depth analyses based on realistic traffic behavior adaptation remain insufficient. Studies on deployment conditions extensively consider factors such as the penetration rates of connected and automated vehicles (CAVs) and automated vehicles (AVs), as well as traffic demand; however, comprehensive studies under more complex traffic compositions and environmental constraints are still limited. Research on lane management and vehicle control lacks systematic analysis and validation in terms of strategy adaptability and responses to traffic disturbances under mixed traffic conditions. In addition, simulation platforms still need further improvement in scenario realism, behavioral modeling accuracy, and unified validation standards. These findings indicate that future research should focus on strengthening the study of dedicated lane deployment under multi-factor coupling conditions, improving adaptive lane management and vehicle control methods for realistic mixed traffic environments, and developing high-fidelity simulation and testing platforms based on virtual-real fusion.

To address the issue of low travel efficiency caused by insufficient bus network coverage in rural areas, this study proposed a bidirectional coordinated scheduling method for demand-responsive transit (DRT). This method integrated the existing fixed-route resources of urban-rural buses and town-village buses with dynamic transfer services. Firstly, the whole-day bidirectional trips of urban-rural buses were divided into rolling time periods. The locations and number of transfer points were determined based on the actual passenger flow conditions in each time period. Meanwhile, the deadheading distance and vehicle repositioning issues during the switching of up and down tasks of town-village buses were systematically considered to achieve effective task connections within each time period. Secondly, to tackle the problem of insufficient service capacity during peak operating hours, a stop response rule was proposed. This rule coordinated the responses to multiple passengers and implemented differentiated pricing strategies for peak periods. Thirdly, a joint scheduling model is constructed with the objective of maximizing system profit. The model set constraints from the perspectives of passengers, urban-rural buses, and town-village buses respectively. Additionally, a four-stage collaborative optimization algorithm guided by spatial clustering of stops was designed. Finally, taking Line No. 4 of the urban-rural bus service in Siyang County, Suqian City, Jiangsu Province as an example, numerical experiments were carried out under different passenger flow intensities. The research results show that, under high passenger flow conditions, compared with the existing strategy, the scheduling strategy proposed in this study increases the number of DRT stops served by 36.84% and the number of DRT passengers transported by 43.48%. Compared with the solution strategy guided by the greedy algorithm, the solution strategy guided by spatial clustering of stops can increase fare revenue by 4% and reduce the driving distance in the demand-responsive section by 4.47%. The dynamic transfer mode significantly reduces the average in-vehicle time of rural DRT passengers by 7.62 minutes. Moreover, the number of rural DRT vehicles participating in the service has a high impact on system profit, accounting for 69.36%. Furthermore, the coordination of multi-passenger responses and peak-period fares can effectively improve the system's ride-sharing rate.

Personalized driving is a pivotal approach to enhancing user trust and optimizing the autonomous driving experience. At present, although Large Language Model (LLM)-based planning and control methods have made significant progress in personalization, they still face challenges such as insufficient generalization, unstable policy iteration, and performance degradation. To address these challenges, this paper proposed a Large Language Model (LLM)-based evolutionary planning and control method for personalized autonomous driving. The method was driven by two core components: ‘style reasoning annotation’ and the ‘coupling of imitation and reinforcement learning’. First, leveraging the logical reasoning capabilities of Qwen2.5-14B, deep mining and style assignment were performed on NGSIM highway data to construct a large-scale driving style dataset. Second, a unified framework coupling imitation learning (IL) and reinforcement learning (RL) was established. Specifically, multi-style Generative Adversarial Imitation Learning (GAIL) was employed to obtain reward functions and base policies that reflected real-world driving data distributions. This addressed the exploration challenges in early-stage policy training and effectively mitigated the negative impact of the “sim-to-real gap.” Subsequently, by incorporating individual rewards refined by takeover feedback, personalized policy iteration was completed within a Proximal Policy Optimization (PPO) framework. During policy updates, a trustworthy evolution mechanism was introduced to regulate the update direction. This mechanism prevented performance degradation, ensured safe policy optimization, and guaranteed continuous improvement of the driving policy. The proposed method was extensively evaluated through open-loop tests on NGSIM and closed-loop simulations on the SUMO-CARLA platform. Experimental results demonstrate that the proposed method possesses the capability of continuous improvement and style adaptation. Compared to classical reinforcement learning algorithms, our approach increases the success rate by at least 5.3% under congested highway traffic; and achieves at least a 3.7% improvement in success rate in emergency scenarios. Compared with existing SOTA planning-and-control methods, our approach improves the success rate by at least 4.9% in congested highway scenarios and by at least 3.5% in urban unprotected left-turning scenarios.

Accurate prediction of the energy consumption of Fuel Cell Vehicle (FCV) serves as a critical foundation for driving range estimation and energy management optimization. To address the adaptability issues of traditional energy consumption prediction models under complex operating conditions, this paper conducts an in-depth analysis of the factors influencing FCV energy consumption and proposes a multi-feature energy consumption prediction model that integrates real-time vehicle operating conditions and environmental parameters. Firstly, a fuel-cell hybrid electric vehicle model is developed based on the AVL CRUISE platform. Through an analysis of vehicle energy flow under multiple operating conditions, the influence mechanisms of features such as speed, acceleration, road gradient, temperature, and humidity on FCV energy consumption are quantitatively elucidated. Subsequently, a multi-dimensional feature vector incorporating real-time vehicle operating conditions and environmental parameters is constructed, based on which a comparative validation framework of “multi-feature vs. multi-model” is designed to train and optimally evaluate energy consumption prediction algorithms. Finally, real vehicle tests verify the generalization capability and accuracy of the multi-feature fusion prediction model. The results demonstrate that the prediction accuracy of the energy consumption model integrating vehicle operating conditions and environmental parameters is significantly improved. With speed, acceleration, road gradient, temperature, and humidity as input features, the energy consumption prediction accuracy reaches 97.2%, representing a 3.7% improvement compared to traditional integrated models that only consider vehicle-side parameters. The research findings provide theoretical support for accurate FCV energy consumption prediction and energy management, holding significant value for engineering applications.

To address the co-optimization challenges of energy management and speed planning for fuel cell heavy-duty trucks (FCHDT) in complex traffic environments, this paper proposes a hierarchical predictive cooperative optimization control strategy (PV-DTD) for intelligent connected vehicle (ICV) scenarios. Through dynamic interaction between the upper and lower layers, real-time coordination of speed planning and energy management is achieved to reduce overall energy consumption. In the upper layer, a Long Short-Term Memory (LSTM)-based preceding-vehicle speed predictor is integrated with a receding-horizon Model Predictive Control (MPC) algorithm to dynamically generate energy-efficient speed trajectories satisfying safety, fuel economy, and ride comfort requirements. In the lower layer, a Dual Time-Domain Adaptive Equivalent Consumption Minimization Strategy (DTD-AECMS) is proposed, in which a neural network (NN) rapidly and accurately estimates the equivalent factor, which is subsequently adjusted dynamically based on the planned future speed profile and real-time State of Charge (SOC) feedback to achieve near-optimal power split between the fuel cell and the traction bat-tery. A full-vehicle simulation platform is constructed in MATLAB/Simulink to evaluate the proposed strategy with respect to speed pre-diction accuracy, speed planning effectiveness, and energy management performance. Results demonstrate that the PV-DTD strategy effectively reduces energy consumption while maintaining driving safety and comfort, achieving hydrogen consumption reductions of 1.65%, 0.81%, and 2.13% relative to the PV-SOC, PV-NNSOC, and HV-DTD strategies, respectively, thereby validating its effective-ness and practical potential in dynamic traffic scenarios.