Automated pavement crack repair offers a promising approach to significantly extend road lifespan and is crucial for intelligent road maintenance. To tackle challenges associated with real-time crack trajectory extraction and substantial sealing errors, the Automated Pavement Crack Sealing Robot (APCSbot) was developed. APCSbot integrates a real-time crack trajectory segmentation network (S2TNet) and a cross-entropy-based adaptive fuzzy control method (CEAFC) for crack sealing repair. The S2TNet incorporates Anchor Ratio IoU Sampling (ARIS) and Balanced Fine-Grained Features (BFGF) to enhance the detector's capability in predicting bounding boxes and segmenting instance binary masks, consequently improving crack trajectory extraction accuracy. The CEAFC method employs cross-entropy optimization iterations to tune controller parameters and constructs fuzzy logic to enhance repair control robustness. Furthermore, an unmanned wheeled robot framework based on four-wheel independent differential drive was established, integrating the crack segmentation network and tracking repair control methods. Extensive experiments conducted on DeepCrack, CFD, and S2T-Crack datasets demonstrate a real-time pavement crack segmentation accuracy of 80.21%. The crack sealing repair process achieves a speed of approximately 0.05 m·s^-1, with an average sealing error for slender cracks of 5.17 mm. The APCSbot showcases its accuracy and robustness in pavement crack sealing repair, thus providing technical support for intelligent road maintenance.

In order to realize the high value and diversified utilization of muddy slag, a new artificial granulation technology that used alkali-activated blast furnace slag (GGBS) was developed to solidify muddy slag and achieve high strength and high water resistance under standard curing conditions. This paper experimentally explored the effects of factors such as alkali activator type, dosage and curing time on the mechanical strength and water resistance of solidified soil particles, and determined Ca(OH)₂ and Na₂SiO₃ as alkali activators and the optimal mixing ratio. On this basis, Fourier transform infrared spectroscopy (FTIR) and scanning electron microscopy (SEM) were further used to microscopically characterize the artificial particles, revealing the strength formation mechanism and evolution law of alkali-activated GGBS solidified soil particles. The study shows that GGBS undergoes hydration reaction under the alkali activation of Na₂SiO₃; when the Na₂SiO₃ dosage reaches more than 10%, the particle strength reaches more than 5 MPa after 7 days of standard curing, and the softening coefficient is higher than 0.75. By adding Ca(OH)₂ to replace part of Na₂SiO₃, the alkali activation effect can be further enhanced, and the optimal mixing ratio is 1∶3. The particle strength is increased by more than 20% compared with the use of Na₂SiO₃ alone. Microscopic experiments show that when alkali-excited GGBS generates hydrated calcium silicate (C—S—H) and hydrated calcium aluminosilicate (C—A—S—H), an inorganic material M_x{—(SiO₂)_zAlO₂—}_n·wH₂O with a high degree of polymerization and a three-dimensional network structure is synthesized, which can fill pores and bond soil particles, significantly enhancing the strength of the soil particle blank. In this paper, artificial aggregate is made by alkali-activated GGBS solidified sludge, which has the characteristics of light weight (1.8-2.0 g·cm^-3), high strength (≥5 MPa), water resistance (softening coefficient＞0.75), low energy consumption (20 ℃ cold curing), green and environmental protection (solid waste utilization rate ≥85%), etc. It can be used to replace natural fillers in traffic roadbed base, backfill of cross-sea bridge pedestals, and protective structures of coastal highways.

The rapid construction of infrastructure will inevitably lead to large-scale maintenance. To address the issues of maintaining the structural performance and enhancing the bearing capacity of long-span bridges, based on a summary of methods for enhancement, as well as closely related technologies such as structural damage simulation, optimization algorithms, and process safety monitoring, this paper focuses on the current status and development trends of improving bridge performance by changing the structural system, systematically reviews the research status and typical engineering applications of enhancement methods such as the transformation of simple-supported into continuous structures, adding support points, cable-stayed composite systems, suspension-composite systems, hanging systems, and beam-arch composite systems, identifies two main types: the method of increasing constraints and the method of additional structures, deeply analyzes the active transformation behavior of the structural system of long-span bridges, as well as the scientific mechanisms behind changes in structural states. The paper outlines the key issues, major challenges, and future development trends related to structural system modification and reinforcement. It highlights that this method involves both benefits and risks, and points out that the compatibility between the old and new systems requires systematic and in-depth research. The establishment of mathematical models for bridge damage remains a shortcoming limiting the research on structural system modification and reinforcement. Issues such as the construction of multi-variable and multi-objective functions and the formulation of optimal solution criteria still need exploration, and breakthroughs in solving algorithms for complex stress processes remain key. With the help of smart devices and digital twins, interactive collaborative design and intelligent construction methods that involve real-time monitoring, analysis, control, and feedback should be further explored.

To thoroughly investigate the fatigue performance of the orthotropic steel-UHPC composite bridge decks with large-size U-ribs, and to address the issues of scale fragmentation and information barriers inherent in traditional full-scale model fatigue tests, a multiscale integrated experimental research method was proposed by taking the interrelationships of performance indicators among the component, subassembly, and structural scales as the starting point. This method adopted a strategy of progressive advancement, integration, and feedback of multiscale performance indicator information, designed multiscale coordinated fatigue tests for components, subassemblies, and structures, established a modular experimental research pathway, and provided effective support for multiscale experimental studies under the same research objective. The results indicate that the proposed multiscale integrated experimental research method, through the information transfer and integration pathways among components, subassemblies, and structures, forms a systematically closed-loop research framework, which can effectively resolve the problem of data fragmentation in multiscale testing. At the component scale, typical fatigue-prone details such as the UHPC material, stud connectors, and steel bridge deck all exhibit performance degradation and fatigue damage accumulation characteristics. These can be quantitatively characterized through mechanical performance degradation models, S-N curves, and damage accumulation criteria, serving as fundamental inputs for upper-scale assessments. At the subassembly scale, segment model tests can elucidate the fatigue damage evolution paths of each segment model, determine their fatigue failure modes, identify the controlling locations of the UHPC layer, stud connectors, and typical fatigue-prone details of the steel bridge deck, and establish mapping relationships between local responses and component performance indicators. This provides experimental basis and theoretical support for design parameter optimization and structural-scale fatigue response analysis. At the structural scale, in-situ monitoring results of the actual bridge show that the strain responses at key controlling locations are stable and the degree of damage is low. The performance indicators demonstrate good applicability and consistency across the component, subassembly, and structural scales.

During the construction of curved sections in super-large cross-section shield tunnels, eccentric jack loads can easily cause joint opening and offset deformation in segment joints. This subsequently induces the degradation of joint waterproofing performance. To address this problem, a complete research framework was established in this study based on a multiple sealing gasket waterproofing system ranging from the “local waterproofing mechanism” to the “global load response”. A fluid-solid coupling analysis model for joint waterproofing performance was set up, and a mechanical analysis model for multi-ring segments under complex construction loads was also established. The study systematically revealed the deformation characteristics of segment joints in super-large shield tunnels caused by eccentric jack loads as well as the degradation pattern of the waterproofing performance of multiple sealing gaskets. The results show that the multiple sealing gasket system exhibits a “gradient barrier and functional synergy” waterproofing mechanism.The waterproofing performance of the three-gasket system is improved by 23.6%~35.3% compared to the double-gasket system, and by 62.5%~76.2% compared to the single-gasket system. The waterproofing performance degradation of the circumferential joint under eccentric loads reaches 21.0%~24.0%. This degradation is most significant when the shield machine adopts an upward attitude. Finally, the degradation level of the waterproofing performance of multiple sealing gaskets in segment joints caused by eccentric jack loads was quantified. A waterproofing safety factor correction method based on the degradation rate of waterproofing performance has been established, and the adjustment values of the corresponding waterproofing safety factors considering the influence of eccentric loads have been clarified. The research can provide theoretical support for the refined design of segment joint waterproofing in shield tunnels with super-large cross-sections under ultra-high water pressure, and provide a scientific basis for shifting the safety factor of joint waterproofing performance from “empirical judgment” to “data-driven”.

This paper proposed Risk Explanation Ability Constructed Technology, an architecture designed to elicit human-like driving risk reasoning capabilities in lightweight pre-trained large language models (LLMs). The method aimed to promote their application in driving risk segment recall and automated analysis of risk causes. This method consisted of a pre-training phase and an iterative optimization phase. In the pre-training phase, a chain-of-thought (CoT) is designed according to the reasoning framework of driving risk, namely risk factor identification, interaction behavior inference, and potential risk determination. A lightweight LLM is then guided using a few-shot learning approach to generate this CoT, enabling it to initially assess driving risk levels and generate corresponding reasoning. In the iterative optimization phase, a guided learning strategy is employed. During the initial optimization stages, a “teacher” model is used to regenerate reasoning for samples with incorrect driving risk level assessments. Correct samples and regenerated samples are collected to conduct supervised fine-tuning. Experiments were conducted using the LLAMA 3-8B model as the base model and Qwen2-72B as the “teacher” model, with 7 000 naturalistic driving segments. The results show that this method improve the risk level assessment accuracy of the lightweight pre-trained model from 0.527 to 0.783. Furthermore, by comparing the similarity between manually constructed risk reasoning and model-generated reasoning, this method improves the ROUGE-L (Recall-Oriented Understudy for Gisting Evaluation-Longest Common Subsequence) metric from 0.517 to 0.616 compared to the baseline model. These results indicate that the proposed method effectively enhances the consistency between the model's reasoning and human risk reasoning. This method provides a feasible approach to automatically analyze the causes of risk, supporting the creation of driver safety profiles and the delivery of targeted safety education.

To enhance decision-making safety in highway obstacle avoidance and overtaking scenarios, and to address the limitations of existing Deep Reinforcement Learning (DRL) methods, which rely on short-term observations and lack trajectory prediction for surrounding traffic participants. This paper proposes a DRL-based driving decision-making method integrated with trajectory prediction information. First, an interactive trajectory prediction module based on a Spatio-temporal Transformer is constructed, which incorporates a spatial attention mechanism and a temporal convolutional network to extract multi-vehicle interaction features and predict the future trajectories of surrounding vehicles. Combined with these prediction results, a dynamic driving risk field is established to achieve a quantitative evaluation of potential collision risks and long-horizon driving safety. Subsequently, a DRL driving decision-making framework integrated with trajectory prediction is designed. This framework explicitly introduces predicted trajectories into the state space and utilizes long short-term memory networks to extract temporal features for optimizing the Actor-Critic architecture. Concurrently, a safety reward function is constructed based on the dynamic driving risk field, and an interpretable safety constraint mechanism is introduced to further ensure decision-making safety. Finally, experiments are conducted using the CARLA simulation platform and a self-developed Hardware-in-the-Loop testbench. The results demonstrate that the proposed Trust Region Policy Optimization with Trajectory Prediction information (TRPO-P) algorithm improves safety and traffic efficiency by 14.80% and 6.39%, respectively, compared to baseline reinforcement learning algorithms. These findings verify the effectiveness of the proposed method in enhancing vehicle safety and traffic efficiency within complex dynamic highway driving scenarios.

Unmanned mining trucks, as the primary carriers for transportation in mining areas, have seen rapid development in recent years. However, due to their large size and numerous blind spots, these trucks are often equipped with multiple LiDARs for surround perception. Achieving high-precision calibration of multiple LiDARs on unmanned mining trucks is crucial for efficient autonomous driving perception. In light of this, this paper proposes a joint self-calibration algorithm for multiple LiDARs on unmanned mining trucks based on a coarse-to-fine calibration (CTFC) approach. Firstly, to address the issue of uneven terrain in unstructured environments, a site usability validation algorithm is proposed, ensuring the primary usability of the input data stream. Secondly, to tackle the problem of inconsistent point cloud sparsity and significant differences in overlapping regions among heterogeneous LiDARs, a multi-LiDAR registration algorithm based on iterative hierarchical reorganization is designed. This algorithm improves joint registration accuracy by extracting identity constraints and aligning the data from coarse to fine multiple times. Finally, to address the weak constraints of non-overlapping LiDAR calibration, a non-overlapping registration algorithm based on bilateral equal-distance constraints is proposed. This algorithm constructs calibration relationships between non-overlapping LiDARs by assuming the identity of calibration board positions observed by multiple LiDARs with overlapping regions. To validate the effectiveness of the proposed algorithm, experiments were conducted in typical feature-degraded scenarios, selecting multiple mining area scenes. The performance of the proposed algorithm was verified based on Root Mean Square Error (Root Mean Square Error, RMSE) and center point matching error metrics. The experimental results show that the proposed algorithm positively impacts the final outcomes. In typical degraded scenarios, the RMSE for multi-LiDAR calibration was 0.048 m, and the center point matching error was 0.028 m. The overall efficiency improved by 120 times compared to manual calibration and multi-stage calibration methods, demonstrating significant advantages.

Environmental perception, as the core technology of the autonomous driving system, directly affects the decision-making level and driving safety of intelligent vehicles. It is the key to achieving high-level autonomous driving for intelligent vehicles. To enhance 3D object detection accuracy and robustness in complex scenarios, aimed at the limitations of the lack of image edge semantics and the interference of point cloud background noise in the current BEV multimodal fusion perception, this paper proposed DDL-BEV, a multi-scale dynamic fusion perception framework based on DepthEdgeNet, Dynamic Queries, and LiDAR-Camera Cross Attention. First, DepthEdgeNet was constructed. The fusion of depth information and edge semantic features was achieved through dual-branch feature extraction and interaction, and the camera Bev space features were optimized. Second, a Dynamic Query module was designed. The LiDAR point cloud was voxelized into cylindrical grids and transformed into BEV features. The dynamic perception of the foreground position effectively reduced the interference of background noise. Finally, LiDAR-Camera-Cross-Attention fusion mechanism was designed. Combined with the Feature Enhancement Module of the multi-branch dilated convolution feature enhancement module, a hierarchical feature interaction architecture was constructed. The BEV features of the LiDAR point cloud and the camera BEV features were fused to achieve the complementary advantages of cross-modal features. The fused features were input into the object detection head to obtain 3D object detection results. Experiments on the nuScenes dataset show that the average detection accuracy (mAP) and comprehensive detection score (NDS) of the DDL-BEV fusion algorithm proposed in this paper reach 69.3% and 71.9% respectively. Compared with the baseline BEVFusion method, they are improved by 1.5% and 1.3% respectively. In special scenarios such as at night, on rainy days, during turns, and at intersections, the mAP of DDL-BEV is increased by 6.7%, 5.42%, 5.45%, and 4.65% respectively, and the scene sensitivity is reduced from 14.11% to 8.33%. Results show that the DDL-BEV detection algorithm has stronger detection robustness in scenarios with insufficient lighting, obstructed environments, and rain-fog interference.

With the growing demand for comprehensive perception in vehicle-infrastructure cooperative systems, roadside multi-modal perception has become a key approach to overcoming the limitations of onboard sensing. This paper proposes an adaptive balanced optimization framework for multi-modal perception guided by a vision-language model (VLM) to enhance the performance of roadside sensing systems. The framework introduces a dynamic weight allocation module that achieves spatially adaptive multi-modal fusion through cross-modal attention and frame-level residual modeling. To address the convergence imbalance among modalities, a gradient-sensitive asynchronous optimizer is designed to finely regulate modality-specific learning rates. In addition, a lightweight gated scheduling mechanism dynamically triggers VLM calibration based on modality states and scene semantic entropy, thereby reducing computational overhead. Experimental results demonstrate that the proposed method achieves 3D object detection mAPs of 79.20% and 80.16% on the DAIR-V2X-I and RCooper datasets, respectively, outperforming comparable methods by an average of 3.9% (up to 7.51%). Meanwhile, the gated scheduling mechanism reduces the average VLM invocation frequency by 41.2%, effectively cutting redundant computation, while the overall GPU memory usage increases by only about 4.0% compared with the baseline. This work provides a novel, efficient, and scalable solution for advancing intelligent perception in vehicle-infrastructure cooperative systems.

To address the limitations in risk perception and analysis methods for complex highway environments, this study introduces a Gaussian risk pulse waveform to quantify risk energy. By combining the energy interaction characteristics of the field, it achieves a theoretical fusion of risk pulses and driving risk fields. Firstly, by defining the maximum/minimum influence ranges of the risk field and considering the underlying constraints of vehicle geometry on the risk interaction space, a dynamic risk field is constructed in combination with vehicle motion characteristic variables, enabling the spatial expression of “behavioral constraints”. Secondly, a static risk field model is built based on functional differences in road markings (boundary, solid, and dashed lines) to quantify the spatial effects of “rule constraints”. Then, utilizing the superimposition of risk energy, a unified risk field perception model with dynamic-static dual-field coupling is constructed. Finally, the ET-SSE algorithm is employed to calculate the risk field strength threshold and classify risk levels. Empirical data from the Kunming-Mohan and Gongxiao highways in Yunnan are selected to validate the model in complex traffic flow scenarios. Comparisons with Time-to-Collision Inverse (TTCI), Time Headway Inverse (THWI), Artificial Neural Network (ANN), Spatial-Temporal deep learning (ST-Transformer), deep reinforcement learning (DRL), and traditional Driving Risk Field (DRF) models show that the proposed unified risk field model considering the risk pulse effect (RP-DRF) exhibits excellent performance in stability, accuracy, continuity, and real-time perception capabilities, with an average risk perception accuracy of 92.77%. Based on the empirical results, risk levels are classified into three categories: low [50, 443), medium ［443, 1 537], and high (1 537, 2 500］. Furthermore, sensitivity analysis reveals the field strength evolution mechanism. This research provides a quantified safety posture tool for risk perception, decision-making, and control of intelligent connected vehicles in highway scenarios, and the established risk level mechanism can be applied to early warning strategy formulation.

In autonomous driving tasks, the main challenge faced by visual large language models is how to perceive the surrounding world and handle complex tasks. However, currently available open-source visual large language models have not been specifically trained during the pre-training stage, resulting in weak spatial understanding and perception capabilities, making them difficult to be directly applied to trajectory planning tasks. In this paper, a dual-enhanced end-to-end trajectory planning framework featuring “spatial question-answering fine-tuning+BEV perception input” is proposed to address the trajectory planning task. Firstly, the visual large language model is trained to recognize different obstacles and spatial messages encountered in autonomous driving based on the annotations of the dataset. Subsequently, bird eye view images are generated from the surround-view cameras to reconstruct spatial information. Finally, the bird eye view, surround-view cameras, and text prompt are input into the spatial enhanced visual large language model. The model is trained through question-and-answer pairs to obtain trajectory data in a standardized format. The effectiveness of the method was verified on the nuScenes dataset and the NAVSIM dataset in this paper. The test results demonstrate that this method has excellent trajectory planning capabilities in real-world scenarios, is more in line with the driving habits of real human drivers, and has generalization capabilities across multiple scenarios.

Safety remains a core challenge in automotive engineering. In recent years, autonomous driving technology has demonstrated potential for improving traffic efficiency and driving safety. However, challenges persist, such as insufficient generalization to long-tail scenarios. Within the current connected driving environment, operational risks primarily stem from control command failures and information uncertainty. Existing safety control methods were often based on single-environment assumptions and they had difficulty handling complex scenarios with overlapping risk. In this work, a driving safety assurance framework based on a safety sandbox was proposed. The framework monitored control commands from autonomous driving algorithms, cloud instructions or human drivers in real time in a non-intrusive way. The system first performs risk assessment based on the current operational state. Then, a resilient arbitration module integrates historical arbitration outputs with the current control commands, applying a multi-level arbitration strategy and resilient intervention mechanism to ensure driving safety while preserving the original driving intent as much as possible. A scaled vehicle-road-cloud co-simulation platform and a real-vehicle testing platform were built. Experimental results show that the proposed method ensures safety under complex conditions, including static and dynamic obstacles. It also maximizes driving comfort. The method provides a feasible and efficient solution for safety assurance in intelligent and connected vehicles.

Compliance with traffic regulations by automated driving systems (ADS) is not only a fundamental prerequisite for ensuring the safe operation of automated vehicles but also a critical factor in enabling automated vehicles to adopt socially integrated driving strategies. However, researchers have yet reached a consensus on the technical requirements, improvement strategies, and evaluation methodologies for ensuring the traffic regulation compliance of ADS. To address this issue, this study systematically reviewed the current research status and challenges regarding the traffic regulation compliance of ADS. First, after clarifying the definition of traffic regulation compliance of ADS, this study presented specific requirements for such compliance and reviewed the existing traffic regulations applicable to both traditional and automated driving. It then conducted a comparative statistical analysis, using a real-world dataset, of common scenarios where traffic regulations were violated by human-driven vehicles versus automated vehicles. Second, a framework for enhancing the traffic regulation compliance of ADS was proposed, comprising four layers: ① a temporal logic-based method for formalizing traffic regulations; ② perception optimization for traffic signal recognition in ADS; ③ vehicle trajectory prediction under traffic regulation constraints; and ④ compliance decision-making in ADS. Research progress on enhancement strategies was reviewed from these four perspectives. Subsequently, regarding the evaluation of traffic regulation compliance of ADS, solutions to three key challenges were summarized: diverse evaluation metrics, feasible experimental methodologies, and the determination of evaluation thresholds. Finally, the challenges faced by the traffic regulation compliance of ADS and future development trends were analyzed and discussed. The obtained results indicate that the traffic regulation compliance of ADS is the underlying support for promoting the large-scale deployment of automated vehicles. To advance this field, future research should focus on several key breakthroughs: establishing a dynamic processing mechanism for hard and soft traffic regulations, designing a dynamic regulation update mechanism integrating natural language processing and knowledge graphs, constructing hierarchical formal models, building a fusion decision-making framework of reinforcement learning and game theory, developing an evaluation system for the traffic regulation compliance of ADS, and innovating the generation of compliance trap scenarios as well as the construction of cross-regional knowledge bases.

Aiming at the energy-saving cooperative control problem of connected vehicles, this paper proposes a hierarchical distributed model predictive control (DMPC) method to achieve multi-objective optimization of both tracking performance and fuel economy. Firstly, a state-space model for individual vehicle longitudinal dynamics and a power polynomial-based fuel consumption model are established. Considering two information flow topologies, predecessor-leader following and two-predecessor following, a complete connected vehicle system model is formulated. On this basis, a hierarchical DMPC controller architecture is designed. The upper-layer controller aims for optimal tracking performance to derive the optimal tracking control input, while the lower-layer controller focuses on minimizing fuel consumption to obtain the optimal economic control input. A novel constrained-domain construction method is introduced to coordinate these objectives, ensuring system stability while enhancing economic efficiency. Design conditions for guaranteeing platoon stability are derived and rigorously proven. Finally, simulation results under sinusoidal disturbance and emergency braking conditions demonstrate that the proposed scheme reduces fuel consumption by 2.92% and 14.75%, respectively, compared to the benchmark scheme, while maintaining satisfactory tracking performance. Overall performance improvements of 4.27% and 14.44% are achieved in the respective scenarios. This study confirms the superiority of the proposed method in terms of energy saving and system stability, providing theoretical support for energy-efficient control of connected vehicles.

Snow and ice on the road in winter will seriously affect traffic safety and transportation efficiency. It is of great significance to clarify the low-temperature freezing behavior of droplets on the surface of pavement materials for the snow and ice control on pavements. Based on a low-temperature adhesion observation system, this study investigated the influence of ambient temperatures, droplet volumes, and substrate surface conditions on the freezing behavior of adhered liquid (H₂O) on cement concrete surfaces. The result indicated that on the cement concrete surface, the freezing of droplets is mainly controlled by heat conduction, and the freezing process consists of super-cooling stage, phase change stage, papillation stage, and completion stage, based on imageology. During the freezing process, the freezing surface in droplets gradually moves upward from the heat conduction interface, while the volume expands with the frozen undergoing, then finally releases in the form of papillations. According to experiments, as the ambient temperature decreases, between 0 ℃ and -4 ℃, the droplets continue to remain in supercooled state without freezing; when the ambient temperature is lower than -4 ℃, the droplets gradually freeze, and the freezing completion time shortens as the ambient temperature decreases. For the original pavement surface, when the ambient temperature is -15 ℃, the freezing completion time is 39.95% less than that at -8 ℃ and only 6.34% less than that at -12 ℃. The increase of liquid volume affects the heat transfer efficiency of the droplets and prolongs the final freezing time. The freezing completion time of 0.5 mL droplet is 8.40% longer than that of 0.3 mL droplet, and the freezing completion time of 0.8 mL droplet is 44.37% longer than that of 0.5 mL droplet. To the initial surface under -12 ℃, affected by the contact area in heat conduction, the freezing time of the droplet is negatively correlated with the surface roughness. The greater the height variance of the concrete surface micro-structure, the faster the droplet freezing process is and the shorter the freezing completion time is. Compared with the normal concrete surface, the freezing completion time of the sandpaper polished surface is extended by 14.13%-16.90%. For pavement surfaces in cold regions, it is appropriate to achieve a balanced design of surface texture depth considering anti-skid and freezing resistance.

To address the challenge of recycling waste tires and to mitigate the problem of differential settlement in fill-cut roadbeds, this study proposed a composite reinforced roadbed system consisting of waste tire cells and geogrids. Through field experiments and theoretical analysis, the settlement evolution and stress distribution characteristics of this technology in fill-cut roadbeds were systematically investigated. Field tests were conducted along the Fenshi Expressway, where the settlement and stress responses of the roadbed were monitored during construction and post-construction stages. The results show that settlement mainly occurs during construction, while post-construction settlement is significantly reduced. Under applied loading, settlement develops more slowly and gradually stabilized. Compared with geogrid reinforced roadbeds, the composite reinforced sections exhibit markedly reduced settlement, and the stress peak was lower than that predicted by the Boussinesq elastic solution. This indicates that the circumferential confinement of waste tire cells and the tensile reinforcement of geogrids work synergistically to achieve lateral load diffusion and improve the uniformity of stress distribution. Furthermore, by treating waste tire cells and geogrids as compressive and tensile units, respectively, a settlement calculation method for the composite reinforced roadbed was established based on the two-parameter elastic foundation beam model. Validation against field measurements confirmed the reliability and applicability of the proposed method. Parametric analysis further revealed that the tensile modulus of tires has a limited effect on settlement, primarily providing lateral confinement; higher tensile stiffness of geogrids more effectively reduced settlement; and the influence of subgrade soil parameters is the most significant, with larger deformation modulus leading to smaller settlement, while higher Poisson's ratio enhances lateral diffusion and improve settlement uniformity.

To promote the development of the reasonable arch axis design theory for arch bridges, the evolution process of this theory was reviewed, the current research status and main problems faced in the calculation methods of reasonable arch axis were summarized, and the future research focuses and directions were discussed. Research results show that the development of arch bridges is intrinsically linked to the improvement of reasonable arch axis design theory. Identifying an arch axis that aligns with the constant load distribution mode and approximates the constant load thrust line is crucial for arch bridge design, which enhances the efficient synergy between material properties and structural force, improving the overall performance and load-bearing efficiency of arch bridges. Calculation methods for reasonable arch axis are generally divided into the analytical equation method and the curve fitting method. Determining the constant load distribution mode of main arch rib and spandrel structures, establishing and solving arch axis equation is the main focus in analytical equation method. Furthermore, selecting the type of curve to fit the arch axis, determining the position and number of control points on the curve, considering curve fitting methods and optimization objectives are the main focus in curve fitting method, the catenary, the high-order parabola and the spline curve are the commonly used fitting curves. The analytical equation method expresses the reasonable arch axis through design parameters of arch bridge, such as constant load intensity and horizontal thrust. This approach provides direct guidance for optimizing the structural configuration of the main arch ribs and spandrel structures, thereby significantly enhancing mechanical performance. In contrast, although the curve fitting method generates geometrically smooth arch axis curves, it inevitably induces substantial local bending moments at concentrated load sections or intermediate sections between adjacent concentrated loads. In order to provide theoretical support for maintaining the reasonable design state of long-span arch bridges, future studies should focus on addressing the design challenges of reasonable arch axis under the synergistic interaction of three scenarios, namely the application of lightweight and high-strength materials, optimization of complex structural layouts, and adaptation of special construction methods. These challenges are in line with the collaborative innovation trends of long-span arch bridges.

To address the issues of false detection, missed detection, poor anti-interference ability and low detection accuracy in existing object detection algorithms during the process of tunnel lining crack detection, this paper proposes a tunnel lining crack detection algorithm RSwin tailored for practical working conditions. The innovation points of this algorithm were: ① It was the first to propose the Residual Swin Transformer Block (RSTB), which had the ability to globally model and locally extract features for complex lining crack characteristics, enhancing the fusion and representation of multi-scale lining crack features and improving the model performance and generalization ability; ② It was the first to integrate the Shape-IoU loss function, optimizing the evaluation method for shape matching problems, comprehensively considering the characteristics of bounding boxes and calculating the loss value based on this, thereby improving the target box matching performance of the model in the task of tunnel lining crack recognition. To verify the effectiveness of the proposed algorithm, a total of 11 classic target detection models (YOLOv7, YOLOv8, YOLOv9, YOLOv10, Cascade Mask R-CNN, Cascade R-CNN, Faster R-CNN, FSAF (Feature Selective Anchor-free Module), FCOS (Fully Convolutional One-stage Object Detection), NAS FCOS (Neural Architecture Search Fully Convolutional One-stage Object Detection), Mask R-CNN) were used on a self-collected tunnel inspection dataset for model comparison, training, validation and testing. The training results and visualization results show that the mAP50 of the RSwin algorithm is 97.6%, which is 14.51%, 5.57%, 4.41%, 2.98%, 3.2%, 2.5%, 6.43%, 11.7%, 3.1%, 4.7%, and 2.4% higher than that of the seven comparison algorithms respectively; at the same time, it has the fastest inference speed, with a frame rate of 9.3 frames·s^-1 under the condition of 807 pixels×606 pixels. The RSwin algorithm has the highest recognition accuracy and the best comprehensive performance, and can be effectively applied to actual tunnel crack detection tasks.

Driven by autonomous driving technology, future roads will inevitably witness a traffic flow pattern of mixed autonomous and human-driven vehicles. In this human-machine mixed environment, the human-machine interaction process requires drivers to adapt, leading to adaptive behaviors in drivers that differ from those in traditional driving environments. Such behavioral changes in turn affect the operation of the entire human-machine mixed traffic flow.This study focuses on data acquisition, behavioral characteristic analysis, and micro-behavior modeling under mixed human-machine traffic flow, reviewing the research status and future prospects of adaptive behaviors in human-driven vehicles. Combining bibliometric methods, this study reviews four mainstream methods of acquiring driving behavior data, analyzes the typical characteristics and influencing factors of drivers' car-following and lane-changing adaptive behaviors in the mixed human-machine traffic flow, and based on this, summarizes three micro-behavior modeling methods for drivers. The research summary reveals that existing data acquisition methods have their own advantages and disadvantages and should be flexibly selected according to different acquisition needs and scenarios. Drivers' adaptive behaviors are mainly reflected in the process of car-following and lane-changing. This behavior is closely related to subjective factors such as drivers' trust in autonomous driving technology and driving style, as well as objective factors such as the penetration rate of autonomous vehicles and the driving environment.Currently, micro-modeling of driver behavior in mixed human-machine environments is still lacking. Existing related models can be categorized into three dimensions: adjusting parameters, introducing human factors indicators, and constructing new models. Based on this, this study further proposes prospective research directions of great academic value, including high-precision multi-modal driving behavior data acquisition, research on adaptive behavior in multi-scenario human-machine interaction, and micro-behavior modeling based on multi-level human factor characterization. These efforts aim to provide a scientific basis for the development of autonomous driving technology, realize harmonious co-driving between humans and machines, and promote the transportation system towards intelligence, safety, and efficiency. aiming to provide a scientific basis for the development of autonomous driving technology, realize harmonious co-driving between humans and machines, and promote the traffic system towards intelligence, safety, and efficiency.

Hybrid electric vehicles exhibit substantial advantages in energy conservation, emission reduction, and alleviation of range anxiety. As the core technology of hybrid systems, energy management strategy (EMS) has emerged as a critical research focus in the field of electric vehicles. With the rapid advancement of machine learning, its capability to process multi-source, high-dimensional data in intelligent connected environments offers novel pathways for optimizing the energy efficiency of electric vehicles. This paper systematically reviews the application and research progress of machine learning in EMS for hybrid systems, with a focus on two primary approaches: intelligent predictive EMS based on classical machine learning and self-learning EMS based on reinforcement learning. Research demonstrates that classical machine learning methods, including supervised and unsupervised learning, effectively extract key features from connected traffic environments and vehicle operational data, playing a crucial role in driving condition preprocessing for intelligent predictive EMS and supporting the optimization of traditional control algorithms. Deep reinforcement learning, which integrates the perception capabilities of deep learning with the decision-making strengths of reinforcement learning, exhibits unique advantages in real-time optimization control for electric vehicles in complex and dynamic traffic scenarios. Building on current research advancements, this paper also provides insights into future directions for machine learning applications in energy management, offering theoretical references for subsequent research.

To explore the distortion-induced fatigue mechanism of welded details under different stiffening ribs configurations and welding types in steel bridge decks, a systematic summary and analysis of domestic and international physical fatigue tests on steel bridge decks and welded stiffening rib details were conducted. The digital twin fatigue test of the welded details between the steel bridge deck and stiffening ribs was carried out, based on numerical fracture mechanics theory and extended finite element method(XFEM). A single-deck steel-truss suspension bridge was selected as the engineering case study. A refined digital twin model of the bridge was established, incorporating 12 typical welded connection configurations between top plates and ribs. The digital twin fatigue test simulated welding thermal effects and realized the complete crack propagation process simulation under the Fatigue Load Model Ⅲ specified in the Chinese code JTG D64—2015. The fundamental concept of crack area expansion rate along with its mathematical formulation is proposed, thus providing a more comprehensive characterization of fatigue crack propagation behavior. Comparative analyses of various structural details were conducted by indicators including loading cycles and crack area expansion rate. The results were cross-validated with physical fatigue tests, achieving data symbiosis and virtual-physical integration within the fatigue metaverse. Physical fatigue test results reveal that: the fatigue strength of the welded toe and root for 8 mm thick and single-sided welded U-ribs is higher than the Category 55 and 50 specified in the Chinese code JTG D64—2015, respectively. For the 8 mm thick double-sided welded U-ribs, no crack initiation was observed at the welded root, and the fatigue strength of the welded toes on the inner and outer sides was classified as Category 70 and 90, respectively. For double-sided welded U-ribs with thickness ranging from 12 mm to 16 mm, the fatigue strength of welded toes on both sides can reach Category 100. The fatigue strength of open-rib welded details exhibits significant variability, but it is generally higher than Category 60. The results of the digital twin fatigue test indicate that, taking crack penetration through the bridge deck as the failure criterion, the fatigue strength of the 8 mm thick and single-sided welded U-ribs corresponds to Category 60 of JTG D64—2015. For the 8 mm thick and double-sided welded U-ribs, the welded toe reaches Category 80. For double-sided welded U-ribs with thickness ranging from 12 mm to 16 mm, the fatigue strength is improved to Category 90. However, the significant welding residual stress and deformation caused by high heat input should not be neglected, as negatively impact the long-term performance of the steel bridge deck. In contrast, the fatigue strength of the open-rib details is only Category 50. The results of the digital twin fatigue test extend the long-life region that cannot be covered by physical fatigue tests, revealing the intrinsic principles of the fatigue metaverse's native data for the steel bridge deck details. These findings provide important insights for analyzing the distortion-induced fatigue mechanisms of steel bridge decks and serve as key evidence for fatigue-resistant design.

As a core component of bridge operation and maintenance (O&M), bridge inspection is increasingly heading towards digitalization and intelligence. Digital twins technology, with its capabilities of virtual-real interaction, data-driven processes, and multi-physical field simulation, has become a critical element in supporting the digitalized and intelligent development of bridge O&M. This study provides a comprehensive review of the technological system for bridge inspection under the digital twin framework. Firstly, a digital twin-integrated framework for bridge inspection were proposed, including data acquisition, processing, and expression. The connotations as well as latest advancements in structural 3D modeling, defect recognition and expression technologies were detailed introduced. Secondly, focusing on the applications of digital twin technology in bridge inspection, the research status and developmental relationships of unmanned autonomous inspection systems and remote real-scene inspection were analyzed. Finally, the key issues including structural digital twin expression, bridge scene data enhancement, and the construction method of digital twin-based bridge inspection systems were discussed, along with prospects for future developments. The review demonstrates that digital twin technology provides a necessary pathway for achieving scientific, refined, and efficient inspection in response to the demand for long-term, comprehensive assessments of bridge structural conditions. While current researches have made significant progress in the development of inspection hardware and software, there still remains substantial scope for foundational theories and technological breakthroughs to be innovated in addressing the critical requirement of O&M under digital twins.

To address the limited perceptual range of physical sensors in the digital twins of engineering structures, a smart virtual sensing technology based on an AI-enhanced reduced-order model (AI-ROM) is proposed. This technology integrates multi-source physical sensor data to predict the global response of complex structures in real time. Methodologically, the structural response is represented as a linear combination of bases using reduced-order modeling theory. This formulation transforms virtual sensing into the optimization of the combination coefficients constrained by the measured response data. Accordingly, a deep-learning loss function capable of handling multi-source physical virtual sensing was constructed. An intelligent model based on a standard attention mechanism was developed for predicting the combination coefficients, enabling the accurate reconstruction of the global structural response using limited local monitoring data. The effectiveness of the proposed method was validated through virtual sensing of a full-scale compression-bending test of the steel-concrete composite tower wall of the Shiziyang Bridge. A refined numerical model was used to generate the full-process response data under combined compression and bending loading conditions to train and deploy the AI-ROM model. In the experiment, measurements from six displacement meters and 17 strain gauges were used as the constraints for response reconstruction. The results show that for highly discrete strain fields, the relative reconstruction error of the AI-ROM is only 9.1%, representing a 63.5% improvement in the accuracy of the refined finite element model. Using the intelligent virtual sensing technology, a sensor optimization method that considers the regional clustering of the analysis domain is further proposed. The spatial distribution of key monitoring points was determined by iteratively evaluating the reconstruction accuracy. In the full-scale tower wall test, this algorithm reduced the number of strain sensors by 58.8%. The proposed intelligent virtual sensing technology facilitates the integration of wide-area perception and real-time simulation in the digital twins of engineering structures, offering more comprehensive information support for safety risk assessments in the operation and maintenance of buildings and infrastructure.

Influenced by complex environmental loads, the mechanical properties of structures evolve throughout their service life. Accurately identifying the characteristic parameters of in-service structures and evaluating their safety performance remain major challenges. This study proposes a hierarchical Bayesian dynamic updating method for time-varying reliability assessment tailored to structural dynamic digital twins. A digital twin model of the physical structure is first established using finite element theory. Based on the monitored dynamic responses of the physical structure, the modal parameters of the digital model are identified through Bayesian inference. A hierarchical Bayesian framework is then constructed to estimate the time-varying physical parameters, whose statistical characteristics are extracted using the expectation-maximization algorithm. The digital model is subsequently updated using the principle of maximum entropy. Leveraging probability density evolution theory, a new approach for evaluating the dynamic reliability of time-varying structures is developed. Compared with traditional methods, the proposed method more comprehensively accounts for the temporal variability of service-related parameters and their associated uncertainties. The effectiveness of the approach is validated through numerical studies on shear-type structures, bridge models, and laboratory experiments, with comparisons drawn against conventional methods. Results demonstrate that the proposed method provides a more accurate acquisition of the parameters of the digital model and significantly improves the handling of parameter variability, thus ensuring the robustness of structural dynamic reliability assessment results.

This study proposed a smart diagnostic method based on Digital Twin (DT) for achieving precise perception and rapid evaluation of in-service bridge safety. Considering structural characteristics and deterioration trends, a “data-driven and knowledge-constrained” DT model for bridge safety assessments was established. The developed model incorporates virtual-real mapping, dynamic interaction, and iterative updating mechanisms. Driven by high resolution 2D images, 3D point cloud data, and other multi-source data, a multi-modal data fusion system was constructed for bridge safety assessments. An object detection algorithm with coordinate attention mechanism and point cloud registration techniques were established to enable rapid and accurate identification and analysis of surface defects and their spatiotemporal evo-lution. By utilizing bridge inspection standards and expert knowledge, an expert knowledge model based on fuzzy logic reasoning was developed for diagnosing intelligent diagnosis of bridge service states, significantly improving the efficiency and accuracy of structure deterioration perceptions and safety assessments. A case study on a bridge in Anhui Province demonstrated that the proposed method could effectively identify defects and deformation conditions, and conduct intelligent diagnosis of bridge safety. The proposed framework is of great significance for advancing intelligent operation and maintenance of bridges.

To develop a low-carbon and sustainable life-cycle maintenance strategy for network-level bridges and enhance the comprehensive benefits in environmental, economic, and safety dimensions, this study proposed an optimization method based on Deep Deterministic Policy Gradient (DDPG) reinforcement learning. This approach integrates bridge structural characteristics, network topology, traffic data, and risk attitudes to construct a reinforcement learning agent that systematically optimizes bridge maintenance decisions. Environmental, economic, and safety indicators were evaluated by comprehensively considering resource consumption, the consequences of potential structural failures, and impacts of vehicle detours, thereby quantifying the contribution of bridge maintenance to sustainability performance. In constructing the reward function for reinforcement learning, sustainability indicators were transformed into monotonically decreasing utility values to reflect the preferences and constraints in the optimization process. Based on a reinforcement learning framework, a DDPG agent with deep neural networks was designed, leveraging the structural degradation features and traffic data of network-level bridges for trial-and-error learning to progressively optimize maintenance decision strategies. The validation results indicated that the reinforcement learning method developed in this study achieves a better balance between environmental, economic, and safety metrics. Through trial-and-error learning, the agent captures the performance variation characteristics of bridges, optimizes maintenance priorities, and allocates resources efficiently. This approach provides a scientific basis for advancing intelligence and sustainability in infrastructure management.

Digital twin and metaverse play a crucial role in realizing the intelligent operation and maintenance of bridges, in which the automated 3D reconstruction of bridges is the key to build the digital twin model and metaverse platform. To address the challenge of integrating geometric and semantic reconstruction in existing 3D bridge reconstruction methods, this study proposed a high-fidelity automated 3D bridge reconstruction method based on 3D Gaussian splatting (3DGS). First, to capture the boundary features of bridge components, a semantic segmentation model for bridges was proposed using the Segment Anything Model (SAM), enabling high-precision semantic segmentation of multi-view UAV-captured bridge images and generating semantic masks for components such as decks and bridge towers. Next, semantic feature attributes were incorporated into the 3D Gaussian kernel functions of scene representation, achieving adaptive characterization of complex geometric and semantic features. Finally, 3D Gaussians were projected into 2D image space, and differentiable rasterization algorithms were utilized to generate semantic feature maps and rendered images. Through joint optimization of geometric, appearance, and semantic attributes, a unified 3D reconstruction was realized, balancing geometric fidelity and semantic accuracy. In this paper, Nansha Bridge in Guangdong Province is used as an example for validation, and the experimental results show that the reconstructed 3D bridge model achieves high-fidelity restoration of texture details and semantic information, with a mean Intersection over Union (mIoU) of 87.2% and an overall accuracy of 91.6%. The proposed method effectively resolves geometric detail loss and semantic fragmentation in traditional approaches, providing a 3D foundational model with both physical precision and semantic interpretability for lifecycle digital twins of bridges.

Surface crack detection in underwater concrete structures is hindered by issues such as low efficiency, high cost, and poor accuracy. This study proposes an intelligent detection method based on deep learning and climbing robots to address these challenges. First, the crack segmentation model Transformer-AL was established based on the crack image features collected by an underwater climbing robot. Subsequently, a feature pyramid learning module was introduced to extract the global semantic information of cracks, and a decoder was designed to match the accuracy of the encoding process. A hybrid-attention-mechanism-based, active-learning module was then proposed to reduce the data annotation cost of the segmentation model. Finally, based on the crack-binarized segmentation results and the calibrated camera, the appearance of underwater concrete structures was reconstructed using geometric quantification and 3D reconstruction techniques. The results show that this method effectively reduces data requirements, achieving a mean intersection-over-union of 0.787 in identifying cracks in complex underwater images. The width quantification accuracy reached the submillimeter level, and the appearance reconstruction results comprehensively displayed the surface cracking condition of the structure. These research findings provide a theoretical foundation for the development of intelligent detection technologies for underwater concrete structures.

To meet the intelligent scheduling and dispatching needs of the smart beam factory under random interference conditions, digital twin technology was introduced to establish a scheduling and dispatching platform that integrated “virtual-physical space interaction, production data analysis, and scheduling decision-making.” Firstly, based on workshop scheduling theory and physical constraints, a framework of scheduling for the smart beam factory was designed, and a no-idle production scheduling model that considered human-machine working hours was established. Then, considering the equipment failure and the component backlog as random disturbances, a dynamic scheduling strategy termed “right shift & partial rescheduling” was proposed. Building on this, the theoretical and technical construction of the digital twin model for the smart beam factory was studied from three aspects: factory elements, production processes, and scenario simulation, and the framework of scheduling for smart beam factory based on digital twin was proposed. Eventually, an application validation was conducted using the scheduling and dispatching of a smart beam factory located along a highway in Sichuan. The results indicate that the proposed no-idle production scheduling model can replace traditional manual scheduling methods, effectively shortening the total production period of the beam factory. The designed dynamic scheduling strategy can effectively respond to unexpected events during the production process. Through static and dynamic scheduling models driven by digital twin, the smart beam factory achieves precise control and dynamic optimization of the component production process. Compared to the static scheduling model, the dynamic scheduling strategy driven by twin data saves 120 and 116 working hours, respectively, under equipment failure and component backlog conditions, resulting in reductions of 14.9% and 14.5%. This research significantly enhances the scheduling flexibility and production efficiency of the smart beam factory.

Subgrade engineering serves as the critical load-bearing component for pavement structures, which significantly influences the stability, safety and durability of road infrastructure. To further advance the sustainability of subgrade engineering in China, propel its high-quality development toward green low-carbon, sustainability, and intelligent development, and contribute to the national strategy of building a “Transportation Power”, this study systematically synthesizes the latest advancements in scientific and technological innovation within China's subgrade engineering domain in recent years, while comprehensively delineating the priority directions for future research. The research is grounded in an analysis of the industry's current development status and evolving trends, and centers on six thematic pillars, namely, engineering properties of subgrade fillers, durable subgrade design theory, subgrade widening technology, subgrade protection and retaining structures, intelligent construction of subgrade engineering, and subgrade disaster prevention and mitigation. Specifically, it encompasses cutting-edge research areas, including the mechanical behaviors of various subgrade soils (e.g., oversized-grained soil, coarse-grained soil, fine-grained soil, and special soils), subgrade moisture evolution mechanisms and associated design methodologies, approaches for determining subgrade structural modulus, calculation and control criteria for subgrade permanent deformation, indices and standards for uneven settlement control of widened subgrade, splicing techniques for new and existing subgrade segments, waterproofing and drainage systems for reconstructed/expanded subgrade, advanced ground improvement technologies, subgrade slope stability assessment, slope protection measures, anti-slide piles and retaining wall structures, design and rehabilitation of slope anchorage systems and waterproof-drainage facilities, intelligent compaction of subgrades, intelligent detection and real-time monitoring of subgrade performance, classification of subgrade disasters, subgrade defect detection, disaster monitoring and warning systems, prevention and mitigation strategies for subgrade defects and disasters, and evaluation and enhancement of subgrade disaster resilience. For each area, this study analyzes and deliberates on the current state of academic research, prevailing challenges, targeted countermeasures, and future development prospects. This review is intended to provide strategic guidance and reference for the advancement of China's subgrade engineering discipline, while offering novel perspectives and foundational insights for researchers and practitioners in this field.

Maintaining an adequate level of vigilance during driving is crucial for driving safety. Types of vigilance decline during driving can be categorized into fatigue, distraction, and prolonged automated driving monitoring. These types may differ in key features and attention mechanisms, but their heterogeneous characteristics remain unclear. This heterogeneity may contribute to the poor generalization ability and suboptimal performance of current models for detecting impaired driving states. This study systematically examines the characteristics and mechanisms of vigilance decline during driving through a literature analysis on measurement tools, types, features, influencing factors, underlying mechanisms, detection methods, and warning systems. The following conclusions were drawn: ① While measurement tools for vigilance have formed a relatively complete system, their application in traffic scenarios is not yet widespread. ② The characteristics of vigilance decline are generally defined, but research on type differences in fatigue driving is insufficient, and studies on EEG features of cognitive distraction are lacking. The effects of auditory-cognitive distraction, fatigue driving, and their interaction on takeover efficiency need further exploration. ③ Mechanistically, vigilance decline due to sleep-related fatigue is linked to reduced cortical activity, while task-related fatigue, distraction, and automated driving monitoring are associated with insufficient attention resources and arousal levels. ④ Existing detection technologies focus excessively on fatigue driving and visual-manual distraction, with insufficient research on detecting cognitive distraction and comprehensive vigilance assessment. The high cost and complexity of EEG and eye-tracking devices limit their use. ⑤ Current warning systems overlook factors such as driving environment and individual physiological and psychological states, lacking differentiated warning strategies based on vigilance decline mechanisms. The following recommendations are proposed: ① Strengthen interdisciplinary collaboration to develop a vigilance measurement paradigm specific to traffic scenarios and promote empirical research on vigilance measurement tools in transportation. ② Systematically compare the vigilance characteristics across different types of fatigue driving and analyze eye-tracking and EEG feature maps of distracted driving under various cognitive processing combinations. ③ Develop portable, low-invasiveness EEG devices and create real-time monitoring models for cognitive distraction based on eye-tracking features and ERP indicators. ④ Overcome ERP identification challenges through standardized experimental design, innovative data analysis methods, and multimodal data fusion techniques. ⑤ Establish classification-based warning standards for drivers, design personalized warning strategies based on vigilance decline mechanisms, and integrate in-vehicle environmental control warning systems.

Traffic signs are crucial elements ensuring the safe, efficient, and green operation of the transportation system. However, the effectiveness of traffic signs is often restricted by the adaptability of traffic sign setting standards, the effectiveness of testing methods, and the comprehensiveness of utility analysis. This paper focuses on the three fundamental issues of “unable to find,” “difficult to understand,” and “incorrect navigation” caused by traffic signs, which affect travel quality. It systematically sorts out and summarizes the research hotspots of traffic signs using the scientific knowledge map method. Meanwhile, starting from the driver's cognitive decision-making chain of “discovery, understanding, and execution,” this paper comprehensively reviews the experimental testing platforms, basic theories, and key technologies in traffic sign research. The study further emphasizes the need to deeply characterize the complex impacts of traffic signs on drivers' visual perception, cognitive processing, manipulative behaviors, and vehicle operating states, and to explore their implicit influence pathways on driving behavior. Additionally, it conducts utility assessments and optimizations of the traffic sign system from the perspective of drivers' information needs. Based on this, combining the team's research experience, this paper innovatively constructs a research and application paradigm for traffic signs covering six aspects: “scheme design, feature representation, quantitative evaluation, optimized selection, supporting arrangements, and standard guidelines.” By analyzing typical traffic sign research cases, it elaborates on the specific implementation steps of this paradigm and compares it with similar research methods at home and abroad. The results show that the research and application paradigm of traffic signs integrating human factor needs has obvious advantages. It not only provides theoretical support for sign design research and optimization but also offers a solid basis for solving common issues related to safety facilities in the transportation industry and significant engineering applications.

Automated vehicles encounter significant safety and efficiency challenges within mixed traffic flows involving frequent pedestrian interactions. Precise modeling of this pedestrian-vehicle interaction is therefore crucial. Such modeling is fundamental not only to advancing the vehicle's decision-making intelligence but also to building high-fidelity virtual testbeds, collectively enabling safer navigation, enhanced user experiences, and superior traffic throughput. Despite its importance, the field currently suffers from highly fragmented research and a lack of systematic review. This paper addresses this gap by providing a comprehensive survey of the state-of-the-art progress and key challenges in pedestrian-vehicle interaction modeling. It first deconstructs the essential characteristics of this interaction through the lenses of traffic safety, utility maximization, social norms, and information exchange to establish a formal definition. It then systematically reviews dominant behavioral modeling techniques and interaction quantification methods, further examining the unique attributes and modeling paradigms of automated vehicles as interactive agents. The paper concludes with a summary and outlook on future technological trends. Our review identifies several critical limitations in the current literature: Theoretical: An inadequate understanding of pedestrian cognitive mechanisms and unsystematic insights into the role of communication; Modeling: The constraints of existing physics- or utility-based assumptions and a lack of research into hybrid models; Contextual: A general disregard for the heterogeneity of interaction scenarios, vehicles, and participants; Methodological: The persistent bottleneck of poor explainability in data-driven approaches. To overcome these challenges, future work must focus on deepening the understanding of cognitive processes, exploring the coupling of physics- and utility-driven models, and systematically integrating contextual factors. Significantly, emerging technologies like multimodal large language models and theories of embodied cognition are creating new research paradigms. We argue that substantial progress in this field necessitates deep interdisciplinary fusion and novel applications of these technologies, paving the way for a next-generation intelligent transportation system that is safer, more efficient, and fundamentally human-centric.

With the rapid increase in the demand for automated driving testing, quickly selecting critical scenarios from numerous testing scenarios has become a top priority. Due to the scarcity and low occurrence probability of critical scenarios, which result in low testing efficiency, there is an urgent need to develop accelerated methods for critical scenario identification. To provide a comprehensive review of accelerated methods for critical scenario identification in automated driving testing, this review is explored in three dimensions: functional scenarios, logical scenarios, and concrete scenarios. In the functional scenario dimension, the research primarily focuses on scenario configuration, selecting the combinations of factors that constitute critical scenarios. In the logical scenario dimension, the research focuses on the scope of scenarios, selecting the range of values for the factors that define critical scenarios. In the concrete scenario dimension, the research emphasizes scenario instances, selecting the specific values of factors that constitute critical scenarios. It is noteworthy that current research on functional and logical scenarios is still insufficient, requiring more scholars to engage in this area. Furthermore, existing methods face multiple challenges, including insufficient scenario authenticity, limited acceleration effects, and incompatibility with automated driving functions. Future research should focus on addressing these issues, particularly in the areas of functional and logical scenarios, and continuously optimizing the accelerating technology for critical scenario identification to provide robust support for the ongoing advancement of automated driving testing technology.

Steel bridges exposed to the atmospheric environment inevitably suffer from corrosion damage under the combined effects of temperature, relative humidity, and pollutants, posing a global challenge. To deepen the understanding of atmospheric corrosion, this study summarizes existing research findings from three perspectives: the macro-environment near bridge sites, the local environment around components, and the micro-environment on component surfaces, while also exploring future research directions. Current research shows that the macro-environment has reached a stage where it can be characterized and classified, along with a classification method based on standard coupon exposure test results and climatic parameters. The local environment only reflects the influence of long-term exposure to corrosive media at different parts of steel bridges but lacks characterization parameters such as the intensity and duration of corrosive media effects, making corrosivity level determination reliant on engineering experience. The micro-environment focuses on the mechanism of steel atmospheric corrosion, dynamically characterizing the corrosivity of different points on steel bridges through parameters such as surface temperature, surface humidity, and surface pollutant deposition. The existing corrosion environment zoning map of China has limitations, including low data density and exposure test stations located far from bridge sites, making it difficult to accurately reflect the corrosivity level of the macro-environment at bridge locations. It is recommended to establish a gradient-based atmospheric corrosion exposure monitoring network along Chinese highway system to obtain multi-source atmospheric corrosion data and develop a corrosion environment zoning map tailored for steel bridges. Future research should aim to build a micro-environment research system that is characterizable, quantifiable, and applicable. Theoretical studies on micro-environment calculations should be conducted to clarify the interaction mechanisms of micro-environment parameters and establish quantitative analysis methods. A predictive model for atmospheric corrosion rates at the micro-environment level should be developed to provide scientific principles for precise detection of localized corrosion, corrosion-resistant structural design, and targeted maintenance strategies.

Driven by global carbon neutrality goals, the clean energy supply for highways presents a critical pathway to decarbonize transportation. This study systematically reviews the characteristics, collection pathways, and utilization potential of green energy in the road area from the perspective of synergistic development between transportation and energy. First, green energy in the road area was categorized into two types based on energy sources: natural energy, such as solar energy, natural wind energy, geothermal energy, and hydro energy; and traffic-induced energy, including mechanical vibration energy, pavement thermal energy, and convective wind energy. Then, the study reviews various energy collection technologies, including photovoltaic cells, wind turbines, heat pumps, hydro and wave energy conversion devices, vibration energy harvesters, and thermoelectric generators, as well as their conversion efficiency and technical challenges. Finally, by establishing a potential assessment model under a unified scenario, the study conducted a comparative analysis of output power, economic viability, and carbon reduction benefits, and summarized typical application scenarios. The study noted that while the potential for green energy is significant, its development and utilization face multi-dimensional challenges, including precise energy assessment, core technology efficiency and durability, and system integration and economic viability. This study aims to provide a theoretical framework and decision-making reference for constructing a clean, low-carbon, efficiently integrated, transportation energy ecosystem.

In order to objectively assess the fatigue damage characteristics of asphalt mixtures under complex service conditions, the strength, fatigue and residual strength tests of asphalt mixtures under different stress states, different test temperatures, and different loading frequencies (rates) have been carried out to reveal the limitations of the traditional fatigue damage model that characterizes fatigue damage with residual strength without taking into account the visco-elasticity characteristics of asphalt mixtures. Based on the three-dimensional strength yield model of asphalt mixtures characterized by effective stress, the fatigue stress ratio under a three-dimensional stress state is defined, the fatigue stress intensity ratio and fatigue life under a three-dimensional stress state are modeled, and the fatigue performance of different stress states, different temperatures, and frequencies is realized to characterize the fatigue performance in a normalized way. Furthermore, a nonlinear fatigue damage evolution model for asphalt mixtures under three-dimensional stress states was derived using the effective stress to characterize the residual strength under different stress states and the normalized fatigue equation to characterize the fatigue life. The results show that the traditional fatigue damage model characterizing fatigue damage by residual strength makes it difficult to objectively characterize the fatigue damage properties of asphalt mixtures under different test methods and conditions, with the model parameter γ₁ fluctuating between 0.933--0.948, and the parameter γ₂ fluctuating between 0.174--0.186. Three-dimensional stress state of asphalt mixture nonlinear fatigue damage evolution model to achieve the fatigue damage of the normalized characterization, not only intuitively verified the fatigue damage of asphalt mixtures of the time-temperature-stress state correlation and equivalence, but also to eliminate the impact of the test method and test conditions on the fatigue damage characterization, for quantitative analysis of asphalt mixtures of fatigue damage characteristics provide a theoretical basis.

Understanding pore water in soils has long been a central and challenging topic in soil mechanics. Its physical properties are also a key scientific issue shared across the mechanical research of various special soils, such as high liquid limit soils and soft soils. Pore water can be classified into free water and adsorptive water, depending on its state. Adsorptive water exhibits unique physical properties, including high density and strong structural characteristics, resulting in its distinct flow and phase change behaviors from free water. However, the effects of the physical properties of adsorptive water on soil permeability, strength, and deformation remain unclear. Moreover, practical engineering generally overlooks the significance of adsorptive water, failing to fully utilize its physical properties to optimize engineering practices. This paper provided a comprehensive review of recent progress on adsorptive water in soil and its influence on soil properties, both domestically and internationally. It covered theoretical and experimental studies across microscopic, mesoscopic, and macroscopic scales. Specifically, it systematically summarizes the formation mechanisms of adsorptive water, analyzes the differences in physical properties between adsorptive water and capillary water, and clarifies the effects of adsorptive water on the hydraulic and mechanical properties of soil. It also reviews experimental research on the effects of adsorptive water on soil permeability, along with recent advancements in permeability coefficient models that consider adsorption. Additionally, the experimental research on the effects of adsorptive water on soil strength is reviewed, along with the development of strength models considering the effects of adsorptive water. Additionally, the paper summarizes the research progress on the role of adsorptive water in soil compression deformation, creep deformation, and subgrade soil resilience, with a focus on its role in the creep behavior of high liquid limit soils and soft soils. Finally, the paper discusses the potential applications of adsorptive water in high liquid limit soil embankments and soft soil foundation engineering, and future research priorities and directions are outlined to provide a reference for further studies.

To reveal the evolutionary law of the mechanical properties of lining structures under the blockage of the longitudinal drainage pipe in karst tunnels, a test device system was independently developed to simulate the blockage of longitudinal drainage pipes and karst channels. The complex tunnel drainage system was equivalent to the circumferential blind pipe and the longitudinal drainage pipe characterized by the cross-sectional area and longitudinal length, and the physical model experiment of tunnel seepage under different blockage degrees of longitudinal drainage pipe was conducted. The results reveal that the groundwater reduction speed in the karst channel and the tunnel drainage volume decrease rapidly with the increase of the blockage of the longitudinal drainage pipe, which significantly reduces the dissipation capacity of the tunnel drainage system. The groundwater in the karst channel directly exerts localized high-water pressure on the tunnel lining, significantly increasing the stress response of the tunnel lining structure. When the longitudinal drainage pipe is completely blocked, the decline rate of groundwater reduction speed and tunnel drainage are as high as 82.5% and 95.9%, respectively. The water pressure at the arch bottom position and the structural stress at the left waist position are the most sensitive, and the growth rates are 30.1% and 37.6%, respectively. Compared with the two blockage indices of the longitudinal drainage pipe, the longitudinal length blockage index directly influences the flow path length, and the cross-sectional area blockage index directly affects the equivalent permeability coefficient, which jointly determine the drainage performance of the tunnel drainage pipe. Furthermore, due to the nonlinear evolution trend, there are noticeable differences in the relative influence weights of two blockage indices during the blockage process, that is, the longitudinal length blockage index is the primary influencing factor under low blockage conditions, whereas the cross-sectional area blockage index becomes the dominant factor under high blockage conditions. The research results can provide a theoretical basis for the safety assessments and maintenance measures of lining structure affected by drainage system blockages during the operational phase of karst tunnels.