Highway construction in China has witnessed remarkable achievements, with rapid growth in the national road network and continuous breakthroughs in the key technologies. This review aims to further enhance the influential level of pavement engineering in China, as well to promote its sustainable and high-quality development. The review systematically summarizes the current status, cutting-edge issues, and future development in pavement engineering. Specifically, it covers seven research topics:highway resilience evaluation and recovery, long-life pavement structures and materials, highway energy self-sufficiency, low environmental impact technologies, the genome of pavement materials and high-throughput computations, highway digitalization and intelligence, and highway intelligent inspection and high-performance maintenance. Focusing on the fields of green, resilience, intelligence, longevity, and traffic-energy interaction, the review identifies 20 critical research topics, including factors leading to highway disasters and their mechanisms evaluation and recovery of highway resilience, key technologies for enhancing highway resilience, full-scale tests for long-life pavement structures, technologies for extending the longevity of highway structures and functions, energy harvesting technologies, energy self-sufficient highways designs, environmental impact testing methodologies and evaluations; innovative materials for low-impact pavements, warm mix asphalt recycling technology, genomic studies on pavement materials, multiscale computation for pavement materials, research on the genome of pavement materials and high-throughput computations, digital modeling technologies, digital twin simulation technologies, data-driven technologies for highway maintenance operations, ground-penetrating radar detection technologies; lightweight detection of pavement performance, strategies for detecting and recovering pavement skid resistance, and high-performance preventive maintenance technologies. The review can provide guidance for the pavement engineering development in China, offering valuable reference for the researchers and practitioners in this field.

The urban multimodal transportation system is a highly complex and diverse transportation network designed to efficiently meet the mobility needs of people, goods, and services within a city. Its complexity originates from many factors including the coupling between different transportation modes, complex interactions between transportation demand and supply, and intrinsic stochasticity and self-organization of an open, heterogeneous, and adaptive system. Therefore, understanding and managing such a complex system is a nontrivial task. However, with the increasing availability of multisource big data in multimodal transportation and other sectors, enhanced computational hardware capabilities, and rapid development of machine learning models, the concept of large models has been applied in various fields, including computer vision and natural language processing. In this study, a conceptual framework, multimodal transportation generative pretrained transformer (MT-GPT), of a data-driven foundation model for multifaceted decision-making in complex multimodal transportation systems was conceived. Considering the characteristics of different transportation modes, the core technologies and their integration methods were investigated to realize this conceptual framework. An expansive data paradigm is envisioned for a foundation model tailored to transportation, along with improvements in hierarchical multitask learning, hierarchical federated learning, hierarchical transfer learning, and hierarchical transformer framework. Application cases of MT-GPT within the “spots-corridors-networks” three-layer large model framework are discussed by constructing “task islands” and “coupling bridges”. MT-GPT aims to provide an intelligent support for tasks such as multiscale multimodal transportation planning, network design, infrastructure construction, and traffic management.

Bridges are crucial infrastructure for traffic and transportation. The inspection of bridge apparent defects is important for ensuring public safety, extending the lifespan of bridges, and identifying risks in a timely manner. They also contribute to improving the reliability and durability of bridges during their operational phases. In recent years, with the rapid development of technologies such as computer vision and artificial intelligence, machine vision has gradually emerged as a new approach for bridge apparent defect inspection. This study conducted a detailed analysis of relevant studies in recent years to review the key techniques for bridge apparent defect inspection based on machine vision, including inspection platform development, data acquisition, image processing, 3D reconstruction, defect localization, and defect parameter quantification techniques. By analyzing the inspection process of existing research, a technical framework for bridge apparent defect inspection based on machine vision was summarized, and the functions and connections between each process were analyzed. The above-mentioned review of key techniques and summary of technical frameworks provide a reference for researchers conducting inspection work on bridge structures. Finally, based on the different levels of automation in data acquisition and defect detection observed in existing studies, this study proposes a hierarchical classification for intelligent bridge apparent defect inspection based on machine vision. This classification includes six levels: manual inspection assistance, defect inspection and localization, partially automated inspection, globally automated inspection, high-degree automated inspection, and fully automated inspection. A comparison of existing literature reveals that although research has moved beyond the traditional stage of manual inspection, it still falls short of achieving fully automated inspection. Therefore, this field has strong research value and broad application prospects.

Decision-making and planning are the core functions of automated driving systems and the key to improving the driving safety, driving experience and travel efficiency of automated vehicles. The main challenges faced by decision-making and planning are how to meet the extremely high reliability and safety requirements for automated driving, and how to effectively deal with scenario complexity, environmental variability, traffic dynamicity, game interactivity, and information completeness, as well as how to generate human-like driving behavior, so that vehicles can integrate into the traffic ecosystem naturally. A systematic and overall review of the technical points of decision-making and planning is presented in this paper to gain a comprehensive understanding of their frontier issues and research progress. Firstly, the research progress of situational awareness-oriented behavior prediction is reviewed from three perspectives, namely data-driven driving behavior prediction, probabilistic model driving behavior prediction, and personalized driving behavior prediction. Secondly, behavior decision-making is summarized into reactive decision-making, learning decision-making and interactive decision-making, all of which are analyzed in turn. Thirdly, motion planning and its applications are compared and analyzed from a methodological perspective, including graph search methods, sampling methods, numerical methods, interpolation and curve fitting methods, etc. Additionally, the key scientific issues and major research progress of end-to-end decision-making and planning are summarized and analyzed. Finally, the significant impact of decision-making and planning on improving the intelligent level of automated vehicles is summarized, and the future development trends and technical challenges are prospected.

Recognition and monitoring of cracks is an important part of the current research on the structural health monitoring of bridges. In the field of inspection and monitoring of bridge structures, traditional crack recognition and monitoring techniques, particularly crack monitoring techniques, hardly meet the timeliness and accuracy requirements of practical projects. Crack recognition based on deep learning has greatly improved the efficiency and accuracy of crack detection; however, it can only obtain crack information at a specific moment, and the ability to monitor the process of crack generation and evolution, which is crucial for a more reasonable evaluation and safety quantification of concrete structures, is lacking. In view of this, a systematic study of crack recognition and monitoring methods based on deep learning was performed. In this study, we analyze and discuss the construction benchmark of a crack dataset, improve and optimize the crack detection and semantic segmentation algorithms, propose a real-time recognition algorithm for multitask integration, establish an evaluation method for the model inference effect, and optimize the calculation method of crack parameters, ultimately forming crack recognition and automatic real-time monitoring algorithms for crack dynamic expansion. The results show that the proposed method for intelligent recognition and monitoring of cracks can effectively track the generation of new cracks and the global evolution of existing cracks, and the monitoring data can provide support for a reasonable and quantitative assessment of the current service performance of bridge structures.

Human-machine collaborative decision-making in intelligent vehicles refers to the process of information sharing, interaction, and collaborative decision-making between the driver and driving automation system. Therefore, it is necessary for enhancing the safety, comfort, and traffic efficiency of intelligent vehicles. Currently, studies on human-machine collaborative decision-making technology are fragmented, and relevant review studies are lacking. Therefore, this study primarily reviewed the current research status and future development trends of human-machine collaborative decision-making technology in intelligent vehicles. First, based on a review of the current research status in human-machine collaborative driving, human-machine collaborative decision-making methods are classified based on driving authority and human-machine interaction. The former includes three decision-making modes:human-centric, human-machine driving authority switching, and machine-centric. The latter comprises two decision-making forms:direct and indirect interactions between humans and machines. Furthermore, based on the hierarchical differences in human-machine collaborative decision-making within intelligent driving systems, human-machine collaborative decision-making technology is divided into two key domains:human-machine collaborative behavior planning and human-machine collaborative motion planning. The former encompasses technologies such as driving-authority arbitration, driving-intent coordination, and human-in-the-loop learning. The latter focuses on two categories of methods:path planning and trajectory planning within human-machine collaboration. Finally, a summary and outlook on the development trends of intelligent vehicle human-machine collaborative decision-making technology are presented. The development of intelligent vehicle human-machine collaborative decision-making technology is not limited to the decision-making level but also relies on common technological progress upstream and downstream, and technologies such as multimodal human-machine interface (HMI) and deep reinforcement learning will become crucial driving forces for its further development. In the future, human-machine collaborative decision-making technology will rely on new decision-making intention transmission technologies and large language models to take it to the next level.

Moisture field is a crucial factor that influences the service performance of a subgrade over its full life cycle. Three issues related to the subgrade moisture field have been extensively examined by the engineering community: the measurement method, evolution law, and control technology. Recently, owing to developments in science and technology, many new theories, methods, and technologies have been proposed for subgrade moisture control. In this study, the current state-of-the-art of the moisture field of the subgrade in the categories of measurement methods, evolution laws, and control technology were introduced. Traditional destructive measurement methods are progressively being replaced by minimal or nondestructive measurement methods, such as time-domain reflectometry (TDR), frequency-domain reflectometry (FDR), and ground penetrating radar (GPR). However, these new measurement methods for subgrade moisture still require improvements for in-situ, nondestructive, precise, and rapid measurements. Theoretical calculations, numerical analyses, and model tests are primarily employed to determine the evolution law in subgrade moisture fields. Currently, the moisture evolution in subgrades has been examined from various aspects, such as unsaturated soil seepage subjected to the hydro-mechanical-thermal coupling effect, flow and heat transfer in porous media, and reduced scaling model tests. Considering the complicated service conditions of subgrades, current research on the moisture evolution law still requires further improvements to fully and precisely describe the moisture field of a subgrade in a real environment. In terms of control technology, subgrade structures with commonly used drainage systems and gravel layers can improve the control performance of subgrade moisture. However, the improvement is gradually weakened with service time owing to the failure to block vapor migration inside the subgrade. Represented by high-performance hydrophobic and hydrophilic materials, new moisture control technology can achieve active control of subgrade moisture. However, more endeavors are still required to improve subgrade moisture control technology with higher precision, higher intelligence, lower cost, higher efficiency, and better practicality.

In order to investigate the stability as well as deformation and failure characteristics of embankment under the combined effect of earthquake and rainfall, this study carried out centrifuge model tests and finite element numerical simulations, where the deformation and failure of embankment subjected to different conditions such as rainfall, post-earthquake rainfall, and post-earthquake rainfall considering the influence of initial rainfall were analyzed. Based on that, the effect of earthquake on the deformation of embankment was examined, after which the development of pore water pressure, evolution of deformation, and characteristics of failure of embankment during post-earthquake rainfall were studied. Results show that earthquake causes the discontinuity in soil deformation and occurrence of surface tension cracks of the embankment. The soil deformation on the sliding surface is most significant and concentrates in the lower part, and decreases gradually with the increase of soil height. Compared with the homogeneous embankment, the pore water pressures of the embankment with earthquake-induced cracks rise more rapidly to the higher level under rainfall after earthquake, and the soil deformation is greatly affected by the cracks, which mainly develops along the cracks near the embankment shoulder, while the deformation of the homogeneous embankment extends to the far position away from the embankment shoulder. The soil saturation of embankment increases and the effective stress decreases as a result of the rainfall-induced wetting effect in the initial stage, thus the deformation of the embankment becomes obvious under earthquake, causing the instability of embankment to be earlier and the deformation and failure more rapid during the post-earthquake rainfall.

In recent years, vessel-bridge collision accidents have occurred frequently. The vessel-bridge collision has become one of the primary causes of bridge failure. To promote the development of bridge collision design methods and anticollision technologies, this paper reviews existing research on vessel-bridge collisions conducted during the past two decades with a focus on risk assessment, response analysis, and protective measures. Firstly, the vessel-bridge collision accidents in China from 2012 to 2022 are investigated to understand the actual collision demand of bridges and fill the information gap on vessel-bridge collision accidents during this period in China. Four new characteristics of vessel-bridge collision accidents in recent years are summarized. The research studies on vessel-bridge collisions are evaluated from the perspective of the adopted approaches. The methods of vessel-bridge collision risk assessment and response calculation are introduced. Four methods are reported for calculating the vessel-bridge collision response, namely, the experimental, contact finite element analysis, equivalent statics, and the equivalent dynamics, from the perspectives of development motivations, characteristics, applications, and limitations. In terms of protective measures, the development history and structural characteristics of two types of passive protective measures, including the floating interception system and the floating anticollision device are summarized. The implementation and updates of the algorithms utilized in active warning protective measures are also discussed. Finally, challenges in current research on vessel-bridge collisions, such as full-factor analysis, multihazard coupling, and standardized design, are identified, and the main directions for future development are outlined.

To address the problems related to the limited perception ability of single sensors and complex late-fusion processing of multi sensors in intelligent vehicle road target detection tasks, this study proposes a multi-modal perception fusion method based on Transformer Cross Attention. First, by utilizing the advantage of cross-attention, which can effectively fuse multimodal information, an end-to-end fusion perception network was constructed to receive the output of visual and point cloud detection networks and perform post-fusion processing. Second, the 3D target detection of the point cloud detection network was subjected to high-recall processing, which was used as an input to the network, along with the target detection output by the visual detector. Finally, the fusion of 2D target information with 3D information was achieved through the network, and the correction of the 3D target detection information was output, yielding more accurate post-fusion detection information. The validation metrics on the KITTI public dataset showed that after introducing 2D detection information through the fusion method proposed in this study, compared with the four benchmark methods, PointPillars, PointRCNN, PV-RCNN, and CenterPoint, the comprehensive average improvements for the three categories of vehicles, cyclists, and pedestrians were 7.07%, 2.82%, 2.46%, and 1.60%, respectively. Compared with rule-based post-fusion methods, the fusion network proposed in this study obtained an average improvement of 1.88% and 4.90% in detecting medium- and highly-difficult samples for pedestrians and cyclists, respectively, indicating that the proposed method has a stronger adaptability and generalization ability. Finally, a real vehicle test platform was constructed, and algorithm validation was performed. A visual qualitative analysis was conducted on selected real vehicle test scenarios, and the detection method and network model proposed in this study were validated under actual road scenarios.

The emerging concept and technology of digital twins provide a key theoretical framework and technical route for the establishment of a traffic system that integrates and interacts with physical and virtual spaces and has received widespread attention from academic, industrial, and government sectors at home and abroad. However, in the process of research and application, there are several pressing issues with the digital twins for traffic systems: ① The concept needs to be clarified; ② The twinning object needs to be defined; ③ The maturity level of the twinning ability needs to be established; ④ Research on new transportation technologies supporting digital twins needs to be strengthened. To this end, this paper systematically sorts out and defines the concept and object of digital twins for traffic systems and further proposes a four-level maturity model called VDMO (“visible-diagnosable-manageable-optimizable”) of the digital twins for traffic systems. Typical application scenarios are taken as the entry point. Then, the research progress, scientific problems, and key technologies corresponding to the four-level maturity model are analyzed. Finally, from the perspective of sustainable development and engineering practice application, the challenges and prospects of digital twins for traffic systems are summarized. This review offers valuable references for technological research, engineering practices, and the sustainable development of digital twins in traffic systems.

In recent years, with increasingly harsh tunnel construction environments and the labor-force population aging, the replacement of human labor with machines has become an inevitable trend in current tunnel construction developments. As one of the two mainstream tunnel construction methods currently employed, the drill-and-blast method possesses greater flexibility and adaptability than the tunnel-boring machine method because of its ability to accommodate various tunnel cross-sectional shapes and geological conditions. Consequently, this method is being more widely applied. Moreover, with the rapid advancement and profound integration of technologies such as big data, Internet of Things, 5G communication, and artificial intelligence, intelligent construction techniques for drilling and blasting methods have witnessed rapid progress. Significant achievements have been made in the intelligent evaluation and blasting design of tunnel surrounding rock, intelligent construction equipment, intelligent construction management platforms, and auxiliary process equipment. Thus, through further integration of information technology, tunnel construction technology, and intelligent equipment, a new model for tunnel construction has emerged, namely, intelligent tunnel construction. This study presents a comprehensive account of the evolutionary trajectory of the drilling and blasting methods in tunnel construction, delving into the intricacies of the current intelligent construction system employed in this method. Subsequently, the notable advancements in intelligent evaluation of the surrounding rock and intelligent blasting design are elucidated. Furthermore, the development and enhancement of intelligent rock-drilling, anchor, arch, wet spraying, and lining trolleys are traced. The current state and application of intelligent control platform technology in tunnel construction is provided. Building upon prior research achievements, this article also discusses the development of tunnel-face surveying and measurement robots as well as collapse-warning robots for tunnel rock masses, with some of these innovations already being successfully implemented on site. Finally, research ideas are proposed to address the challenges of autonomous navigation, control, and operation of intelligent robots in long, large, and deeply buried tunnels, with the aim of providing guidance for the development of tunnel automation technology and proactive disaster prevention and control in China.

With the promotion and application of electric vehicles, the scale of vehicles has exploded. Simultaneously, a series of problems have emerged on the product side, operation side, and service side of electric vehicles:the effectiveness of product quality evaluation is low, the operation safety monitoring is inaccurate, and the service management is incomplete. These problems affect the safe and reliable operation of electric vehicles, restrict the development of the electric vehicle industry, and have become a significant issue for the industry. In the era of automobile digitalization, where big data serves as the core, advancing the development of key technologies in electric vehicle management and services is crucial. This not only ensures the safe, reliable, and efficient application of electric vehicles but also provides essential support and assurance in alleviating consumer safety concerns, ultimately fostering the industry's leapfrog development. To provide a comprehensive overview of technology and innovation in the field, this study extensively discusses the research status of key technologies in electric vehicle management and services. It delves into three aspects:electric vehicle quality assessment and traceability technology, safe operation control and charging service technology, and behavior analysis, identification, and carbon accounting technology. The focus centers on the technical system of 'operation big data +' in the electric vehicle 'quality evaluation-safety monitoring-service management.' Finally, summarizes the existing problems in product quality supervision, operation safety monitoring, and public service management of key technologies of new energy vehicle management and service. Outlook the "quality monitoring-operation control-service management" technology of new energy vehicles with big data analysis and mining as the core. This review can provide a reference for the further improvement of electric vehicle management and service-related technologies and also provide a reference for the future development direction of technology in this field.

To achieve efficient, safe, and accurate quantitative identification of cracks in concrete towers of large bridges, this study proposes a crack recognition and detection method for bridge towers based on wall-climbing robots and transformers. This method uses a wall-climbing robot based on the negative-pressure adhesion mechanism as the equipment carrier to move quickly on a concrete bridge tower and collect high-resolution data from apparent diseases at close range. A crack quantification evaluation method, including image stitching, a Feature Pyramid Transformer (FPT) crack segmentation network, and quantitative dimension calculations, was established. The width data of the key positions of the crack were obtained, along with the visualization result of the global distribution of the crack. Additionally, the proposed method effectively improved the crack segmentation and boundary location accuracy of microcracks. The feasibility and recognition accuracy of the method were verified experimentally, and the method was successfully applied to the main tower crack detection project of a cross-sea bridge in Eastern China. The multi-scale cracks distributed in the width range of 0.1-3.3 mm on the bridge tower surface were detected, and the maximum error of width quantification was less than 10%. The results of this study have good application prospects and are expected to promote the intelligent development of existing concrete bridge structure detection technologies and equipment.

To address the problems of tracking accuracy degradation and stability deterioration when operating intelligent vehicles under changing driving conditions, a multi-objective control strategy based on reinforcement learning variable parameter model predictive control (MPC) algorithm was proposed in this study. The proposed method effectively realizes the parameter adaptive tuning of intelligent vehicle path tracking control system. The proposed linear time-varying MPC controller was designed based on a vehicle dynamics model to obtain the optimal front-wheel steering angle and additional yaw moment. Based on the Actor-Critic reinforcement learning architecture, the Deep Deterministic Policy Gradient (DDPG) and Twin Delayed Deep Deterministic Policy Gradient (TD3) agents were designed for control parameter tuning. The gain function was constructed with tracking accuracy and system stability as the goal, and two typical simulation scenarios of docking road and variable curvature road were constructed for the algorithm performance verification. For the docking road scenario, the prediction horizon and weight matrix of the controller were adjusted in time according to the changes in the road adhesion coefficient. Whereas for the variable curvature road scenario, the prediction horizon and weight matrix of the controller were adjusted in time according to the changes in the road curvature. Through joint simulation analyses conducted using MATLAB/SimuLink, CarSim, and Python, the reinforcement learning-tuned MPC was compared with fixed parameter MPC and Fuzzy-tuned MPC models. The results showed that the reinforcement learning methods yielded the best performance regarding the path tracking accuracy of intelligent vehicles under different road conditions, while guaranteeing the vehicle safety as much as possible. Under the docking road condition, compared with the fixed parameter MPC and Fuzzy-tuned MPC models, the average lateral deviation of the vehicle was reduced by 99.8% and 97.6%, respectively, when using the reinforcement learning-tuned MPC, and the average front-wheel angle change rate was reduced by 99.7% and 77.0%, respectively. Moreover, under the variable curvature road condition, the average lateral deviation was decreased by 79.6% and 90.8%, respectively, and the average front-wheel angle change rate decreased by 40.6% and 2.6%, respectively, compared with those obtained when using the fixed parameter MPC and Fuzzy-tuned MPC models.

Millimeter-wave (MMW) radars are important components of roadside perception systems on smart roads and have been widely used to monitor the status of traffic flow operation, for intelligent control, vehicle-road collaboration, and autonomous driving. However, changes in the relative position and posture of vehicles in traffic and the MMW radar may affect the radar signal echo and point cloud distribution, leading to deviations in the radar perception results of vehicle positions. It is crucial to analyze the spatial distribution characteristics of the perception accuracy of MMW radars to guide their application on smart roads. Based on the perception principle of the MMW radar, this study comprehensively considers the sources of perception errors in the two stages of MMW radar signal processing and point cloud data processing. Through a combination of numerical simulations and field experiments, qualitative and quantitative analyses were conducted on the perception accuracy characteristics of MMW radar under different relative vehicle positions and postures. The results show that the longitudinal perception accuracy of radar is mainly affected by the relative position of the vehicle. When the vehicle and radar are too close (longitudinal distance less than 30 m) or too far (longitudinal distance more than 200 m), the perception result can significantly shift towards the front or rear of the vehicle, and the longitudinal perception error usually exceeds 0.5 m. The lateral perception accuracy of radar is mainly affected by the lateral position and relative posture of the vehicle. When the vehicle deviates too much from the radar center beam (more than 5 m) or the yaw angle is large (more than 40°), the perception result can significantly shift towards the side of the vehicle body and the lateral perception error usually exceeds 0.5 m. Based on the analysis of the influencing factors, this study further provides guidance for the application of MMW radar perception data and the deployment of perception devices on smart roads.

The test specification rationality of the test site directly affects the credibility of vehicle durability testing results. To address the issue of neglecting the priority level of correlated channels when solving the damage-correlated model of the 'user-test site’, an optimization method based on weight analysis of the correlated channels was proposed. A series of sub-objectives were divided according to vehicle correlation requirements and signal types. The weights of the sub-objective functions were analyzed using the criteria importance through intercriteria correlation (CRITIC) approach, and a new comprehensive objective function was constructed by merging the weights using the compromise programming method. Finally, the function was solved using the genetic algorithm. Correlation analysis validation was conducted on a light commercial vehicle to evaluate the effectiveness of the methodologies proposed. Five sub-objects of the vehicle were defined, including wheel center vertical force, longitudinal/lateral force, suspension displacement, force, and strain signals. These sub-objects were assigned weights of 0.255, 0.230, 0.153, 0.203, and 0.159, respectively. Then, the comprehensive objective function was established and subsequently solved to obtain the cycles of the reinforced testing roads. The obtained solution was compared with that from the direct multi-objective solution method. The results indicate that the relative damage ratios of the key channels-wheel center vertical forces match 0.8-1.1, and the rest maintain between 0.5-2, while the relative damage ratios calculated via the direct multi-objective-based method range between 0.4-2.5. This suggests that the proposed method effectively reflects the importance of each correlated sub-target and achieves a more consistent reproduction of vehicle damage under correlation-matching requirements in practical applications. Moreover, the corresponding load distribution and test mileage show that the developed test specification meets the theoretical demands of an enhanced acceleration test in actual engineering. This research provides a valuable reference for vehicle durability testing.

The planning of electric vehicle (EV) charging station has significant impacts on facility construction costs, long-term profitability, and user satisfaction. In this paper, an optimization location model of multilevel charging stations is proposed, which contains two noninterchangeable optimization objectives, minimizing the total cost of charging station construction and vehicle driving cost, to consider the construction cost and service capacity of different levels of charging stations. By analyzing the big data of the existing vehicle trajectories, various vehicle parking states and durations were retrieved, and the potential charging demand of EVs was obtained based on an improved K-means clustering algorithm for the initial location of charging stations. A multilevel EV charging station location model was constructed to analyze the charging demand, which was solved based on the Tabu Search algorithm to evaluate the secondary locations of charging stations. A charging equipment capacity optimization model was established based on the queueing theory. Global Positioning System trajectory data of electric taxis in Chengdu, China were used for an empirical analysis. The charging station locations were determined based on K-means clustering and multilevel location model. The two schemes were compared based on service capacity, economic performance, as well as convenience issues. The location schemes obtained by the two methods are feasible. The coverage of the 2.5 km service radius of the multilevel locations is 91.2%, slightly worse than that of the K-means clustering scheme (93.6%), while both are above 90% coverage rate. In terms of economic benefits, the scheme based on the location model is significantly better than the K-means clustering scheme. The total profit is increased by 14.4%, while the profit rate is increased by 23.4% based on the multilevel modeling scenario with a relative reduction of 11.7% in construction cost. The proposed location and capacity scheme of EV charging facilities can effectively reduce the construction cost of charging facilities, and improve the profitability and user charging convenience, and thus has social benefits.

To realize automatic and intelligent inspection of main cables of a suspension bridge, a route planning method for cable inspection using an unmanned aerial vehicle (UAV) and apparent defect identification with small-sized samples is proposed. First, UAV oblique photogrammetry is utilized to rapidly construct a three-dimensional model of the targeted suspension bridge, facilitating a proposed automatic route planning for UAV inspection of main cables. Subsequently, the Faster R-CNN neural network model is employed to identify apparent defects such as cracks, corrosion, and scratches from images of the main cables. Finally, an image fusion-based data augmentation method is used to improve the accuracy of defect detection with a small-sized sample dataset. During the training process of the Faster R-CNN neural network model, the average accuracy of the three types of defects (i.e., cracks, corrosion, and scratches) in the test dataset increases with the increase of the number of training epochs and gradually stabilizes after the 15th epoch. After 100 training epochs, the average accuracy for the three types of defects in the test dataset reaches 0.723. Field main cable inspections were conducted on the Xiaolongwan Bridge, and the results indicate that automatic UAV route planning for the inspection of main cables based on the established three-dimensional model is feasible in practice. The Faster R-CNN network model can accurately identify cracks, corrosion, and scratches in the main cables. The proposed fusion-based data augmentation method can effectively enhance the defect identification accuracy from small-sized samples.

The rising threat of accidental explosions, terrorist attacks, and precision guided weapon strikes has motivated increasing interest in the anti-explosion safety protection of bridge structures. However, these challenges have received inadequate attention in the current bridge-structure design process in China. This paper summarizes the research development and the main achievements, at home and abroad, of eight aspects: the explosion load model, bridge deck and main girder anti-explosion and protection, pier and cable anti-explosion and protection, bridge underwater anti-explosion and protection, whole bridge structure anti-explosion and protection, bridge explosive vulnerability analysis, and resilience assessment. The main challenges facing these aspects are identified. We then propose the main directions for future research, giving a new perspective and some reference for anti-explosion and safety protection research relating to bridge structures.

To fully leverage the advantages of the Transformer model in high-resolution (HR) bridge crack image segmentation, a refined cascaded segmentation method, Cascade CATransUNet, based on the Transformer and Coordinate Attention (CA) mechanism was proposed. Firstly, a TransUNet-based crack feature extraction module was introduced to preliminarily extract coarse-grained crack features at three scales from low-resolution (LR) crack images. The CA mechanism was incorporated into the skip-connection structure of TransUNet to enhance the representation of subtle crack features. Then, based on the extracted coarse-grained crack features at the three scales, two refined operation modules based on physical cascaded structures were designed to sequentially restore fine-grained pixel features of the crack body and edge region from both global and local dimensions. Additionally, to fully utilize the advantages of multi-scale features in the fine-grained representation of crack boundaries, a multi-scale cascaded loss with an active boundary regression term is introduced during the training process. Ablation experiments conducted on HR bridge crack images captured by the unmanned aerial vehicle (UAV) demonstrated the effectiveness of each proposed component. Finally, the comparative experiment conducted on 4 K-resolution bridge crack images revealed that the Cascade CATransUNet surpasses the state-of-the-art high-resolution (HR) refinement networks Segfix and CascadePSP, both of which rely on traditional convolutional neural networks (CNNs). Notably, the Cascade CATransUNet achieved significant enhancements of 5.04% and 7.10% in mean Intersection over Union (mIoU) and mean Boundary Accuracy (mBA), respectively, while retaining identical GPU memory requirements. By adopting the Cascade CATransUNet, it becomes feasible to perform fine-grained segmentation of HR crack images, enabling structural inspectors to obtain more comprehensive and accurate crack information. Consequently, this provides valuable technical support for bridge safety assessment and maintenance decision-making processes.

End-to-end autonomous driving methodologies eliminate the need for manually defined rules and explicit module interfaces. Instead, these approaches directly map trajectory points or control signals from raw sensor data, thereby addressing the inherent shortcomings associated with traditional modular methods, such as information loss and cascading errors, and overcoming the performance limitations imposed by rule-driven frameworks. Recent advancements in self-supervised-learning-based generative artificial intelligence have exhibited substantial emergent intelligence capabilities, significantly promoting the evolution of end-to-end methodologies. However, the existing literature lacks a comprehensive synthesis of the advancements in generative end-to-end autonomous driving. Consequently, this paper systematically reviews the research progress, technical challenges, and developmental trends in end-to-end autonomous driving. Initially, the input and output modalities of the end-to-end models are delineated. Based on the historical progression of end-to-end autonomous driving, this paper provides an overview and comparative analysis of the foundational concepts, current research status, and technical challenges of traditional, modular, and generative end-to-end methods. Subsequently, the evaluation methodologies and training datasets utilized for end-to-end models are summarized. Furthermore, this paper explores the challenges currently faced by end-to-end autonomous driving technologies in relation to generalization, interpretability, causality, safety, and comfort. Finally, predictions are made for the future trends of end-to-end autonomous driving, emphasizing the fact that edge scenarios provide critical support for the training of end-to-end models, which can enhance the generalization capabilities. In addition, self-supervised learning can effectively improve training efficiency, personalized driving can optimize user experience, and world models represent a pivotal direction for the further advancement of end-to-end autonomous driving. The findings of this research serve as a significant reference for refining the theoretical framework and enhancing the performance of end-to-end autonomous driving systems.

To develop a merging control algorithm for intelligent connected vehicles (ICVs) on freeway acceleration lanes interacting with human-driven vehicles (HDVs) on the mainline, we propose a merging control model (DQN-RF). This model integrates the deep Q-network (DQN) algorithm and the random forest (RF) algorithm. First, a roadside data acquisition platform was established to collect the naturalistic merging behavior data of HDVs at a typical merging zone with an acceleration lane on the G70 freeway in China. Second, a human-like merging decision model using RF was built using historical merging environmental contextual data and the merging urgency of the merging vehicle on the acceleration lane as input. We constructed a simulated merging scenario featuring an acceleration lane on the freeway using the simulation of urban mobility (SUMO) platform. Utilizing the Python language, we developed a testing script environment for the deep reinforcement learning algorithm. Additionally, we introduced a longitudinal acceleration control algorithm based on DQN. Finally, the DQN-RF merging control model, which embedded the RF merging decision algorithm into the DQN longitudinal acceleration control algorithm, was established to embrace merging decision control and longitudinal acceleration control in a comprehensive framework. The default lane-changing control algorithm in SUMO, known as "LC2013," was also combined with the proposed DQN algorithm to serve as a baseline model. The simulation results show that, with the same acceleration action value space[-1, 2] m·s^-2, compared to the DQN-LC2013 model, the DQN-RF model receives a higher total reward value. The average accelerations of the ICV for the DQN-RF and DQN-LC2013 models are 0.55 and 0.09 m·s^-2, respectively. Furthermore, the average speeds are 21.4 and 19.7 m·s^-1, respectively. There are no stop-and-wait phenomena observed when the DQN-RF model is adopted, while there are seven stop-and-wait events in 100 turns of simulation when the DQN-LC2013 model is adopted. The proposed DQN-RF merging control model can realize human-like merging decisions and improve the merging efficiency and success rate of the ICV. The DQN-RF model can be used for merging decision control and longitudinal acceleration control of the ICVs on the freeway acceleration lane.

This paper presented a station-subpath metanetwork-based approach for modeling and solving the optimal charging infrastructure location problem. Specifically, this paper proposed a two-phase mixed linear integer programming model, and accordingly developed a two-phase algorithm powered by the branch-and-bound method, decomposing any individual routing-charging decision into two phases. The first phase aimed to find the distance-constrained minimum-cost subpath between each charging station pair in the original network, which was handled by the bi-criterion label-correcting algorithm; while the main algorithmic step of the second phase was to, in the branch-and-bound framework, repeatedly identified the minimum-cost path between each origin-destination pair in the metanetwork, which can be efficiently solved by the classic single-criterion label-setting algorithm. The numerical results from applying the developed metanetwork-based approach for the Yangtze River Delta network reveal that the construction budget limit of charging stations and driving range limit of electric vehicles play important roles in charging station location decisions and individual route-and-charge choices. When applied to three different sizes of numerical networks, the metanetwork-based approach proposed in this paper exhibits dominantly higher computational efficiency than the conventional network-based approach for this type of problems of large size.

With the increasing occurrence of traffic issues, including congestion, accidents, emission, and energy consumption, it is important to prevent and manage them by improving the capacity and modernization of urban traffic governance. Focusing on urban road traffic, urban road traffic examination (URTE) is a critical foundation and effective section for traffic governance. It aims to identify and diagnose urban road traffic issues timely and accurately by monitoring the health status of traffic in the context of urban examination. With the rapid development of digitization, informatization, and automation, the technology for acquisition and analysis of traffic data has changed. URTE has already become a research focus, but is still in its infancy. In this study, based on an analysis of its connotation, extension, and basic theory, the concept and system structure of URTE are introduced, including the research and application system. According to the research system, the related fundamental theory and key technologies of URTE are reviewed. Relevant traffic data's essential features, application value, and corresponding traffic parameter calculation methods, including model-driven and data-driven, are summarized. Subsequently, characteristics of holistic urban road traffic, identification and corresponding features of traffic issues, and methods of indicator construction are reviewed. Furthermore, the causes of urban road traffic issues were analyzed from traffic planning, design, management, and control perspectives. Although the research on URTE has started to progress, a comprehensive review of this topic has revealed some key challenges that need to be addressed, including the construction of the URTE index system, diagnosis of the traffic health status, and analysis of traffic issue causes. Therefore, future research directions are proposed accordingly to provide a reference for URTE research.

Current road design standards consider the perception reaction time of human drivers and operational characteristics of vehicles. However, there are differences between automated vehicles and human drivers in terms of the perception reaction time, perception distance, and perception height. These differences may lead to challenges for automated vehicles under specific road conditions such as complex road geometries and irregular intersections. To understand the impact of road infrastructure on the operation of automated vehicles, it is necessary to study road readiness for automated driving. Quantitative studies on the effects of different road types, road geometry designs, traffic control facilities, and roadside infrastructure on the safe operation of automated vehicles are insufficient. This paper summarized the findings from three perspectives: road segments, intersections, and traffic signs and markings. The results indicate the following: ① Existing research has established a relationship between road geometry design, intersection geometry design, and the perception capabilities of automated driving. It was found that design parameters such as the horizontal curve radius, chord length, and curve deflection angle affect the sight distance in automated driving. ② Real-world testing showed that the road type, horizontal curve curvature, and lane width have a significant impact on automated driving failure events. ③ An increase in the market penetration of automated vehicles will lead to clearer rules for interaction among automated vehicles and enhanced road readiness. ④ The accuracy of automated driving perception and recognition is affected by the retroreflectivity and placement of traffic signs, whereas the failure probability is related to the size of traffic markings, retroreflectivity coefficient, and extension edge lines. To satisfy the future demands of automated driving applications, it is necessary to conduct tests on existing roads to assess the impact of various complex operating conditions on the operation of automated vehicles. It is necessary to examine the impact of road design on the performance and safety of automated vehicles on new roads.

The existing methods for detecting surface damage on concrete bridges based on semantic segmentation suffer from several drawbacks, including large model parameters, inadequate feature extraction, and low segmentation accuracy. To address these challenges, this study proposes Segformer-SP, a lightweight surface-damage detection method for concrete bridges based on an enhanced Segformer architecture. Segformer-SP adopts the MiT B0 as the encoder and introduces two novel modules: the semantic fusion module (SFM) and polarized self-attention (PSA). In Segformer-SP, the SFM is employed to fuse low-level features with high-level features, thereby enhancing the semantic information captured by the low-level features. Additionally, PSA captures global contextual information during the segmentation process, effectively addressing the issue of insufficient feature representation for damages and further improving the segmentation accuracy. Experimental results demonstrate that Segformer-SP outperforms Segformer-B0, exhibiting a 2.41% increase in mean intersection over union (mIoU) and a 1.91% increase in mean F1-score (mF1). Moreover, Segformer-SP achieves better performance in terms of mIoU and mF1 than most state-of-the-art semantic segmentation models. Notably, Segformer-SP has a significantly reduced parameter count of only 6.09×10⁶, while maintaining a high frame-per-second (FPS) rate of 56.54, making it suitable for deployment on terminal detection equipment.

In a complex traffic environment where autonomous and human-driven vehicles coexist, reducing the influence of complex interactions between two vehicle types, which have drastically different driving behaviors on vehicle driving safety, ride comfort, and traffic efficiency is a key issue that needs to be addressed in the field of autonomous driving decision-making and control. Accordingly, this study proposed a non-cooperative game interaction framework between human-driven vehicles (HV) and autonomous vehicles (AV) in a mixed driving environment. First, a longitudinal game strategy for human-driven vehicles was established, considering the driver's longitudinal control characteristics of linearly decreasing vehicle acceleration, differentiated coordination degree, and different characteristics of time delay. Second, a longitudinal game strategy for autonomous vehicles was designed, considering the safety constraints of autonomous vehicles and surrounding vehicles, as well as the comfort and traffic efficiency objectives constraining the autonomous vehicles during the lane-changing process. Then, the interactions between human-driven vehicles and autonomous vehicles in different mixed-driving environments were solved based on the Stackelberg game theory to obtain the optimal lane-changing gaps and longitudinal speed trajectories of autonomous vehicles. The model predictive control (MPC) method was used to generate safe lateral lane-changing trajectories for autonomous vehicles. Finally, multiple sets of mixed driving conditions were designed according to the differences in the coordination degree and response delay time of human-driven vehicles. The test results showed that autonomous vehicles could quickly and accurately identify the coordination degree of human-driven vehicles, select the optimal lane-changing gap, and cooperate with surrounding autonomous vehicles to merge into the target gap. During the lane-changing process, the autonomous vehicles always maintained a safe distance from the surrounding vehicles, and both the longitudinal and lateral accelerations of the lane-changing vehicle did not exceed 1.25 m·s^-2 at a speed of 20 m·s^-1. Finally, the safety and comfort performance were guaranteed, verifying the effectiveness of the non-cooperative game interaction framework proposed in this study.

Material informatics are the theoretical core of the Materials Genome Initiative, which provides critical methods for performance improvement and new material development of asphalt mixtures composed of multiphase, random structures, and complex behaviors at multiple scales. This paper reviews the application of material informatics in the performance prediction and durability enhancement of bitumen and asphalt mixtures to promote the research and application of material informatics. First, the essential connotations of the material genome and material informatics are analyzed, and their applications in asphaltic materials are summarized. Common material data standards are then summarized. The development of multiscale characteristics and material gene databases for bitumen and asphalt mixtures is reviewed. Furthermore, research on asphalt property prediction and modified asphalt composition optimization based on the chemical composition and colloidal structure genes is introduced. The application of data mining and machine learning algorithms for predicting the mechanical and service performances of asphalt mixtures is outlined, including the mixture design indicators, dynamic modulus, high-temperature rutting resistance, fatigue resistance, low-temperature cracking resistance, and water stability. The composition and structure optimization of asphalt mixtures based on performance prediction and intelligent optimization methods are analyzed to improve the performance of the asphalt mixture. Finally, the framework of the informatics for asphalt mixture materials is discussed. The potential challenges in the asphalt mixture gene system and performance prediction using machine learning are analyzed. Potential problems for future material informatics research are also discussed. This review could provide promotion to the durability improvement of asphalt pavement materials.

To overcome the limitations of existing wall-climbing robots in detecting apparent diseases in high pier and column structures, a climbing inspection operation robot system based on the ring belt vision scanning of high pier and column structures with apparent diseases was proposed. To meet the functional requirements of the climbing robot, a multi-legged cooperative climbing detection robot platform was designed. The operating parameters of the robot operation process can be set by the host computer, and the attitude information of the robot is collected in real time by using multiple inclination sensors installed on the ring frame. The climbing attitude and accuracy were corrected in real time by the designed system algorithm. A ring-shaped track and visual scanning vehicle that can be mounted on the operating platform was designed to carry an industrial camera, a lighting system, and a high-precision encoder to visually scan diseases in the track direction. Finally, the acquired images were used to recognize and measure superficial diseases, based on which a digital archive of superficial diseases was created. Using the constructed prototype, field tests were performed at Guangyun Bridge in Xi'an and Shouchun Bridge in Anhui Province. The results showed that the robotic system can achieve full-area scanning of the pier and column structure, obtain high-precision visual images of the apparent diseases of the pier and column structure, and accurately identify and quantify the diseases to realize their intelligent detection.

Distributed electric drive vehicles can improve vehicle mobility through in-situ steering capabilities. All four wheels of the vehicle are experiencing slipping, creating a situation where the vehicle body is prone to drifting or even losing control. To achieve stable and accurate in-situ steering control, this article analyzed the dynamic mechanism of in-situ steering and proposed a control strategy that coordinates yaw and slip rates. This study designed a road adhesion estimation algorithm based on longitudinal dynamics to complete the evaluation of road conditions before in-situ steering. It adopted a hierarchical control architecture. The upper-layer controller coordinated the torque control strategy based on the vehicle state, and the lower-layer control designed a yaw rate decision-making framework. The nominal yaw rate of the original steering was determined by considering the throttle opening, and the four-wheel-drive torque was calculated using a single-neuron adaptive PID (SNAPID) control algorithm. This algorithm, based on a quadratic performance index, was employed to achieve tracking control of the yaw speed. The expected slip rate of the four wheels was obtained by introducing fuzzy logic reasoning, and the drive torque adjustment was calculated using the PID algorithm and was combined with the yaw rate torque control to suppress the deviation of the steering center. Simulation tests and real vehicle tests show that:when the adhesion coefficient is constant, the tire turning in a steady state is regarded as a rigid body, and the side slip angle and the lateral reaction force on the ground are unchanged, the test results are consistent with the inferred steering dynamics. The test results confirmed that the yaw rate tracking control algorithm exhibits excellent tracking performance and robustness across various attachments. In comparison to PID, the algorithm demonstrated a 46% increase in response speed, a 24.0% reduction in maximum overshoot, and a shortened average adjustment time by 1.3 seconds. The average steady-state errors were consistently maintained at 0.01 (°)·s^-1. The control of yaw rate and slip rate contributed to a reduction in horizontal and vertical coordinate offsets by 2.94 meters and 1.69 meters, effectively mitigating steering center offset.

The purpose of this study was to develop a new application of energy-absorbing materials in the road maintenance field and to produce a preventive maintenance seal that can improve the road surface function and enhance the structural bearing capacity of existing roads. A new road energy-absorbing material was used as the matrix, and a “sandwich” structure was used as the framework. A road maintenance energy-absorbing seal was designed and prepared, considering texture reconstruction. Image-processing analysis and accelerated loading tests were performed to analyze the decay law of the surface texture characteristics and road surface function of the road maintenance energy-absorbing seal and evaluate the durability of the seal. The effect of the road maintenance energy-absorption seal on decreasing the strain at the bottom of an asphalt concrete plate was evaluated using a continuous loading test of wheel rolling, and its load-bearing and buffering effects were investigated. Based on dynamic thermomechanical analysis, the microenergy-absorbing characteristics and damping behavior of the road maintenance energy-absorption seal were described, and its buffering mechanism was revealed. Finally, this study lays a solid foundation for the extensive investigation and promotion of the road maintenance energy-absorbing seal. The results show that the aggregate coverage rate is 40%, based on the seal surface texture and surface functions (wear resistance and sliding resistance). The ratio between the 2.36-4.75 mm and 1.18-2.36 mm aggregates is 25:75. The spraying plans for the energy-absorbing material are 1.0 and 2.0 kg·m^-2 for the upper and lower layers, respectively. After 40 000 cycles of loading and wear cycles, the surface texture of the road maintenance energy-absorbing seal attenuated slightly, and the decline in the durability of its surface was evident. A road maintenance energy-absorbing seal can effectively reduce longitudinal and transverse strains at the bottom of an asphalt concrete plate. Moreover, it can convert the original tensile strain into compressive strain or decrease the value of the original tensile/compressive strain by over 30%-50%. The loss factor [tan(δ)] of the energy-absorbing seal is 0.1-0.3, and the seal can exhibit excellent damping performance within wide ranges of temperature (-50 ℃-200 ℃) and frequency (10^-4-10⁸ Hz).

It is crucial to evaluate the comprehensiveness of safety testing for autonomous driving. Therefore, in this study, evaluation objectives of logical scenario evaluation including comprehensiveness, accuracy, and visibility are proposed. Furthermore, a method based on test results in discrete specific scenarios to identify the danger field in logical scenarios, including danger field modes, distributions, and proportions, is proposed to comprehensively evaluate the safety of the system under test (SUT) at the logical scenario level. First, specific dangerous scenarios were clustered by the Mean Shift algorithm to discover different categories of dangerous scenarios. Second, a Decision Tree by memorization feature selection is proposed to partition the boundary of each danger field mode. Third, the partitioning path was analyzed to automatically calculate the proportion of the danger field. To verify the proposed danger field identification method, it was compared to the baseline on the multimodal test function. The results show that the proposed method is better than the baseline method in the calculation accuracy of the danger field modes, distributions, and proportions. Furthermore, application experiments were conducted on test functions and logical scenario. On the test function, danger field identification was carried out based on different optimization algorithms, verifying the universality of the proposed danger field identification method. The search efficiencies of different optimization methods were compared based on the identification results. In the logical scenarios, three danger field modes with a proportion of 44% and corresponding spatial distributions are identified by the proposed identification method. Based on the analysis of two typical dangerous scenarios, four improvement requirements are proposed for the SUT. The research results show that the proposed method of danger field identification can effectively identify the danger field in logical scenarios of autonomous driving, and comprehensively and accurately evaluate the safety of the SUT at the logical scenario level.

The automated driving technology still faces many safety challenges in mixed traffic environments. Intersections are high-risk locations for autonomous vehicles (AVs). The aims of this study are pre-crash scenario generation and crash causation analysis of AVs. A pre-crash scenario method of roadway-traffic participant-critical event-precrash movement was developed. Thirty one pre-crash scenarios at intersections for AVs were generated using 470 crash reports involving AVs in California. Significant differences between the AV and conventional vehicle (CV) pre-crash scenarios were verified by a statistical analysis. A crash causation method is proposed based on the system control structure, which reveals the interaction relationship between AV crashes and roadway, traffic situation, environment, automated driving system, driver (tester), and vehicle. Nine crash causation patterns and causation chains of AV crashes in the rear-end and lane change scenarios were determined. The results indicate the following: AVs being rear-ended by CVs occurred with a frequency 4.03 times that of rear-ended CVs. The main reasons for rear-end scenarios were that the driver of a CV follows the lead vehicle too closely and insufficient decision-making of the automated driving system to decelerate first, and then stop or start. The main reasons for lane change scenarios were dangerous lane changes or overtaking of CVs, insufficient recognition of other vehicles' lane change intentions by the automated driving system, and unreasonable decision-making of deceleration and collision avoidance. This study promotes the application of scenario-based crash causation analysis methods. It can guide the construction of automated driving test scenarios, and provide a reference for the development and optimization of automated driving systems and improvement in intersection safety.

To evaluate the fire resistance of hanger systems in suspension bridges, vehicle fires are classified into five levels. Levels 1 and 2 represent passenger vehicle fires, levels 3 and 4 correspond to truck fires, and level 5 represents tanker fires. These vehicle fires are characterized by distinct maximum heat release rates and burning durations. The proposed hierarchy was validated using existing vehicle fire experiments. Geometric features of flames are established for Levels 3, 4, and 5 vehicle fires based on previous vehicle fire incidents. For passenger vehicle fires, a cylindrical flame radiation model was employed to compute spatial radiative heat flux, validated through three full-scale car fire tests. In the case of truck fires, a prismatic flame radiation model was used to calculate spatial radiative heat flux. For tanker fires with crosswinds, a computational fluid dynamics method validated by a liquefied natural gas trench fire test was employed to calculate the heat flux envelope on the hanger surface. An incremental temperature calculation formula for hanger cross-sections with radiative heat flux boundary conditions was derived, and validation was performed using finite element models. Using mechanical property tests of high-strength steel wires at high temperatures, a quantitative relationship between critical temperature and design safety factor of hangers is developed based on the ultimate load-carrying capacity at high temperatures. Finally, integrating the above outcomes, a five-step theoretical framework is proposed to evaluate the fire resistance of hanger systems under graded vehicle fires. This algorithm can serve as a reference for the assessment and fire-resistance design of hanger systems in suspension bridges.

Superabsorbent polymer (SAP) has been developed as an optimum internal curing material for concrete owing to its superior absorption, storage, and water release behavior. To better understand the relationship between SAP water absorption-release behavior and concrete performance and increase the curing efficiency of SAP, three theories of SAP water absorption-release were summarized. The water absorption-release behavior in different solutions, water exchange mechanism between SAP and cement paste, and relationship between SAP and concrete internal humidity were systematically analyzed. The test methods of SAP water migration behavior were summarized, and the characteristics and applicability of each method were analyzed. The effects of the SAP particle size, dosage, and entrained water amount on concrete properties were discussed from the perspectives of cement hydration, pore structure, and micro-morphology. The results show that the ionic network, solution thermodynamic, and gel phase transition theories can explain the water absorption-release behavior of SAP in different solutions, and the amount of adsorbed and capillary water in SAP affects the water release time of SAP. The water absorption-release properties of the SAP determine the moisture distribution in concrete. The microstructure and macroscopic properties of SAP internal curing concrete are comprehensively influenced by SAP dosage, particle size, and entrained water amount. Appropriate parameters enable hydration products to fill the pores created by SAP, thus refining the concrete pore structure and improving compactness, mechanical properties, and durability. Additionally, the self-healing capacity, resistance to fire spalling, and algae cultivation of concrete are increased by SAP. To improve the design of the internal curing concrete of SAP, the morphological evolution of SAP and its water absorption-release behavior in concrete, as well as the relationship between SAP and water distribution, should be clarified.

A method used in computer vision-based bolt loosening detection involves correcting the perspective of connection node images and analyzing changes in the straight-line angles of the bolt contour edges. However, this method still has significant room for improvement in the reliability of image correction and bolt contour edge detection. Therefore, a visual detection method based on deep learning is proposed for detecting bolt loosening in the bolted nodes of steel structures. First, the bolts were detected using YOLOv5. Next, for the bolt region of interest (RoI), the contour edge straight lines of the bolts were extracted using Attention U-Net with high robustness. To improve the bolt object detection accuracy, the object detection model should set a lower confidence threshold to ensure that there are no misdetected objects and then screen out false bolts based on the number of edge straight lines extracted from the bolt RoI. The node image was corrected using the perspective transformation method. The reference points required for the transformation were automatically positioned according to the intersection relationship between the bolt detection boxes after spatial movement. Finally, the bolt angle was calculated based on the straight lines at the edge of the bolt contour in the corrected image, and loosening was determined by detecting the angle difference between the detection and reference bolts. The results show that: the AP value of bolt object detection is 0.97; the mean values of accuracy, recall, and F1 of bolt contour edge detection are 0.846, 0.807, and 0.825, respectively, with high robustness under a variety of complex background interferences; the false bolt screening method can filter out 99.82% of the false bolt objects; the proposed image correction method applies to connection nodes with a variety of common bolt arrangements; when the loosening discrimination threshold is only 2.8°, the accuracy of bolt loosening detection is up to 99.7%. This method has good application prospects for the automated operation and maintenance of bolted nodes in large-scale steel structures.

Underwater pile-pier structures are important components of bridges. Various surface defects occur on these structures due to their complex hydrological environment. Existing methods for the visual detection of such defects have two main issues: ① Underwater images are blurred, and the colors are severely distorted; ② the size of defects cannot be quantitatively identified, and the detection efficiency is low. To solve these problems, this paper proposes a method to extract the contours of underwater pile-pier surface defects by combining an image fusion enhancement algorithm with a deep learning model. First, a pixel-level image fusion algorithm based on point sharpness weights is used, which can fuse two single enhanced images as well as significantly improve image contrast while ensuring effective color correction. Second, the DeepLabv3+ semantic segmentation network model is improved in terms of weight, such that the number of weight parameters required for the model can be reduced as much as possible while maintaining the accuracy. Next, an open-source dataset of surface defects in building structures is used to train the backbone feature extraction network layer, and the transfer learning method is applied to the detection task of the object domain. Finally, the image dataset collected from underwater experiments and practical engineering works is used to train the light-weight improved model, establish the underwater pile-pier surface defect contour extraction model, and then verify and test the models. In addition, comparisons focusing on three aspects, namely, comparison with five other commonly used algorithms, comparison of detection results with and without image fusion, and comparison with and without noise effects, are made to verify the robustness and effectiveness of the proposed method. The results show that the image fusion enhancement algorithm proposed in this paper can effectively enhance the detailed features of the defect image contours, and the light-weight improved model has the highest recognition accuracy and can maintain high detection efficiency and robustness. This implies that the proposed method is suitable for the quantitative detection of the surface defect contours of underwater pile-pier structures implanted in small underwater robots for practical bridge structures.

The truck-drone delivery mothership system refers to a joint delivery pattern in which a truck carries drones to locations close to customers, launches these drones to serve multiple customers, and then retrieves the drones. This is an important development direction with potential in the field of traffic engineering. Owing to the demands of some customers related to the load capacity of a drone and locations outside the flight range of a drone, a truck-drone joint delivery system that considers customers with greater demands and at greater distances (TDJD-CGDGD) based on the mothership system is proposed, where the truck is allowed to serve customers in question and drones can be retrieved at different locations from where they are launched. This delivery pattern can be viewed as a traveling salesman problem for drones. An MILP model aimed at minimizing the total delivery cost was formulated. To solve large-scale instances efficiently, an algorithm hybridizing the greedy randomized adaptive search procedure (GRASP) and adaptive large neighborhood search (ALNS) was developed. This algorithm first routes trucks and drones with an additional constraint that can simplify truck-drone simultaneous routing. This additional constraint is then relaxed, and the algorithm focuses on adjusting the drone routes to further reduce the total cost. It was found that our algorithm had good performance; TDJD-CGDGD achieved an average cost saving of 19% compared to truck-only delivery, allowing drones to be launched and retrieved to service customers with high demands, resulting in an average cost saving of 5% compared to not allowing this function.

Fatigue cracking significantly affects the service performance of orthotropic steel bridge deck (OSD), making their reinforcement and treatment the focus of intense research in recent years. This paper analyzes and summarizes the development of fatigue cracking reinforcement for OSDs, focusing on key issues and future development directions. The results indicate that multiple fatigue cracking modes of the welded joints are the fundamental attributes of fatigue problems in OSDs. The applicability and effectiveness of reinforcement methods are closely related to the fatigue cracking mode and the fatigue crack length. Taking assembly reinforcement as an example, a new synergistic structure system is developed by incorporating reinforced components in existing structures. Local stress characteristics of the fatigue cracking area in the synergistic structure system is changed notably in contrast with the existing structure, thereby achieving performance enhancement and longevity. The reinforcement system contains existing fatigue cracks, cumulative fatigue damage in uncracked modes, and parts vulnerable to new fatigue generated by reinforcement. The fatigue cracking modes of the reinforcement system are diverse and complicated. Key issues such as the damage status of each cracking mode before reinforcement, the stress state of the reinforcement system, the cracking mode of the reinforcement system, and the prediction of its remaining lifetime are important foundations for studying OSD reinforcement and treatment. The assessment of the actual damage status and damage reconstruction methods for in-service OSDs, the remaining lifetime prediction of the synergistic structure system, as well as the reinforcement design based on the remaining lifetime of the reinforcement system are issues of considerable interest for fatigue cracking reinforcement in OSDs. The reinforcement method based on new materials and the treatment methods applicable to OSDs with significant fatigue cracking are important research topics that require urgent attention.