31 March 2026, Volume 39 Issue 3
    

  • Select all
    |
    Special Column on Perception, Decision-making and Control for Intelligent Connected Vehicles
  • Multi-LiDAR Self-calibration Algorithm for Mining Trucks Based on Hierarchical Registration
    WANG Jie, YANG Song-yue, YU Gui-zhen, WANG Zhang-yu, LIU Run-sen, ZHANG Shuai, WANG Ji-fu
    China Journal of Highway and Transport. 2026, 39(3): 1-18. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.001
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Research on Depth-enhanced BEV Perception Technology in Traffic Scenarios Based on Fisheye Camera
    XU Qi-min, LIU Yang, Lü Jia-xuan, ZHANG Zhi
    China Journal of Highway and Transport. 2026, 39(3): 19-33. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.002
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Bird's Eye View (BEV)-based perception provides a more comprehensive dynamic semantic map for autonomous driving by offering an expanded field of view. However, fisheye cameras, while benefiting from a wide field of view, also suffer from significant distortion issues, which in turn leads to lower accuracy when applied to BEV perception. To address this issue, this paper proposes a BEV perception method for fisheye cameras based on depth information enhancement. Firstly, a feature extraction network more suitable for fisheye images is established by taking the visual Transformer model as the basic architecture and designing spherical windows, position embeddings, and multi-head attention calculation modules. Then, a depth estimation model network architecture is proposed to improve the depth estimation accuracy of fisheye images. The semantic information of the image is introduced into the decoding layer of the depth estimation network to guide the depth prediction process. Finally, by combining the depth estimation of fisheye images with the Transformer model, the predicted depth information is used to construct depth query vectors through a depth information enhancement module, guiding the Transformer decoder to achieve the mapping from column features to BEV features. The BEV semantic map is ultimately output through a semantic segmentation network. Experimental results indicate that on the SynWoodscape dataset, the proposed method improves the mIoU of BEV perception to 50.7%, surpassing the baseline model HFT by 4.8%, thereby validating its effectiveness in the fisheye image BEV perception task.
  • Dust Segmentation Based on Concentration Perception for Vehicular Vision in Unstructured Scenes
    FU Yao, GAO Ming, XIE Guo-tao, LIU Qun-xin, BIAN You-gang
    China Journal of Highway and Transport. 2026, 39(3): 34-49. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.003
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    The pervasive dust in unstructured environments can severely degrade the perceptual performance of intelligent vehicle vision systems. To address this issue, this study proposed a dust segmentation method based on a Dust Concentration-sensing Nested U-structured Convolutional Network (DCSNet), aimed at achieving fine-grained segmentation of dusty regions in images. Firstly, this research proposed a Ghost Convolutional Channel Spatial Attention (GCSA) module, which integrated depthwise separable-Ghost convolution, multi-semantic spatial attention, graph channel attention, and residual connections, enabling more efficient feature extraction. Secondly, a difference amplification module was designed to filter the predicted dust probabilities, suppressing pixels with low probabilities and enhancing the expression of pixels with high probabilities. It also mapped similar probability values to a broader range, thereby enhancing the network segmentation capabilities. Thirdly, in response to the inability of current dust visualizations to represent concentration change, this study employed predicted probabilities to characterize dust concentration, resulting in a smoother transition between dust edges and the background. Lastly, considering the lack of open-source real-world datasets for unstructured scenarios, which were mostly generated through simulation tools, this study constructed a dataset comprising dust images from real open-pit mining environments under various lighting, weather, and operational conditions, providing data support for algorithm training and evaluation. Experimental results demonstrate that the performance of DCSNet in dust segmentation surpasses various state-of-the-art methods, thereby validating the superiority of this method in dust segmentation tasks.
  • Research on Dynamic Defogging and Multi-object Detection Methods Based on YOLOv8-MRDE
    JIANG Shu-xia, WU Jie, ZHOU Yong-jun, CUI Xiang-bo, HUANG Cheng-xiang, GONG Gui-liang
    China Journal of Highway and Transport. 2026, 39(3): 50-61. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.004
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To address the challenges of multi-target detection accuracy degradation and robustness deterioration caused by insufficient vehicle environmental perception under adverse weather conditions, this paper proposes an enhanced YOLOv8-MRDE algorithm to improve target detection performance in foggy driving scenarios. The algorithm achieves performance optimization through three key improvements: First, a MixDehaze dehazing module is integrated at the front-end of the backbone network to effectively enhance image feature visibility. Second, a multi-scale feature fusion architecture based on Reparameterized Generalized-FPN(RepGFPN)is constructed, which strengthens cross-scale feature representation through hierarchical feature reproduction mechanisms. Third, a dynamic attention mechanism is introduced in the detection head to establish a Dynamic Head(DyHead) structure for improved capture of critical features. For training optimization, the EIoU loss function replaces the conventional CIoU to accelerate network convergence, while structural pruning techniques are applied to eliminate model redundancy and achieve lightweight deployment. Experimental results on both the RTTS real-world fog dataset and Foggy Cityscapes dataset demonstrate the superior detection accuracy of YOLOv8-MRDE. Compared to the baseline YOLOv8 model, the proposed algorithm achieves mean Average Precision (mAP) improvements of 2.6% (RTTS) and 4.2% (Foggy Cityscapes), with 25% and 22% reductions in model parameters and computational costs, respectively. The findings validate the effectiveness of YOLOv8-MRDE in foggy driving conditions, demonstrating its potential to enhance detection accuracy and robustness. This work provides both theoretical foundations and technical support for improving safety in low-visibility autonomous driving systems.
  • Research on LiDAR/Camera Fusion Perception Method for Autonomous Driving Based on DDL-BEV
    CHEN Jing-jing, SHANGGUAN Wei, CHEN Jun-jie, CHAI Lin-guo, QIU Wei-zhi
    China Journal of Highway and Transport. 2026, 39(3): 62-74. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.005
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Multi-modal Iterative Optimization Fusion Algorithm for Multi-target Tracking of Unmanned Vehicles
    LI Jie, WANG Shi-min, WANG Xiao-yan, ZHOU Wei-jia, WANG Chang-cheng, CUI Ya-feng, HU Zheng, LAN Hai, WANG Zhi-yong
    China Journal of Highway and Transport. 2026, 39(3): 75-87. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.006
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    In order to enhance the environment sensing ability and multi-target tracking accuracy of a car during driving, multi-sensor fusion is used to make up for the shortcomings of a single sensor. However, the accumulation of these errors will affect the accuracy of the fusion result. In order to ensure the tracking performance of multi-target tracking and the accuracy of multi-modal fusion, it is necessary to solve the problems of motion modelling, association matching and multi-modal fusion. In this paper, we use multi-category motion models to predict the future trajectory, and we introduce the CTRA and Bicycle models into the motion models, and we use the different object categories to select different motion models to accurately express different types of motion. In real scenes, different categories of objects often exhibit different geometric features, and the problem of 3D target association matching needs to be solved. Therefore, this paper designs multi-stage data association and incorporates the BIoU cost metric to cope with the problem of association matching of point clouds in 3D space. In the fusion stage, adaptive weight iterative trackless Kalman filter fusion algorithm is used, and the fusion weights are iterated repeatedly to get the optimal weights. The computed results are evaluated and compared with other typical tracking algorithms. Compared with the optimal algorithms in other typical tracking methods, the average multi-target tracking accuracy (AMOTA) has increased by 1.83%, and the total number of identity switching (IDS) has decreased by 4.54%. It is demonstrated that the method proposed in this paper can improve the environment perception accuracy and has better tracking stability and robustness in multiple categories of scenarios.
  • Vision-language Model Guided Adaptive Balanced Optimization Framework for Roadside Multi-modal Perception
    ZHANG Guo-yu, CHEN Qian, SUN Jian, HANG Peng
    China Journal of Highway and Transport. 2026, 39(3): 88-100. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.007
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Cooperative Perception Enhancement of Intelligent Connected Vehicles Under the Pose Error Interference
    ZHA Yuan-yuan, SHANGGUAN Wei, CHEN Jun-jie, CHAI Lin-guo
    China Journal of Highway and Transport. 2026, 39(3): 101-115. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.008
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To meet the perception requirements of intelligent connected vehicles in complex traffic environments, a cooperative perception enhancement method based on pose error calibration is proposed to address the issues of limited single-vehicle perception field of view, inconsistent availability of connected interaction data, and inaccurate pose information of multi-source cooperation. Focusing on the cooperation effectiveness, the data availability, and the perception precision, we designed a distributed architecture of cooperative perception enhancement for intelligent connected vehicles, including the determination of cooperative agents, the screening of interaction data, and the calculation of cooperative perception. The Transformer cross-attention mechanism was used to fuse the LiDAR point cloud and image to optimize the autonomous perception ability of connected traffic agents. With the constraint of communication distance, the cooperative graph model was constructed based on the principle of perception enhancement in the safety-critical area of the ego vehicle and perception complementarity among cooperative agents to determine the agents for cooperation. Then, to avoid the influence of abnormal perception results, spatial topological consistency and perceived result continuity were used to filter interactive data. Finally, the pose error calibration was completed through the association matching of high-availability perception interaction data. The redundant perception results and new perception results are cooperated based on the cooperative gain calculation and object visibility distinction, respectively. Cooperative perception enhancement of V2X multi-source interaction data was achieved. To verify the effectiveness of the proposed method, tests and verifications were conducted based on the OPV2V dataset and the V2X-Real dataset. The experimental results show that compared with single-vehicle perception, the cooperative perception enhancement method proposed in this paper improves the AP50 by 22.2% and 21.7% on the OPV2V and V2X-Real datasets, respectively. Compared with mainstream cooperative perception methods, the method proposed in this paper has the smallest drop value of AP50 under the pose error interference. Moreover, for the OPV2V and V2X-Real datasets, the reduction value of AP50 is only 48.4% and 52.3% of that of the RobustV2V method, respectively. It was verified that the proposed method improves perception precision while reducing the interference of pose errors on cooperative perception performance.
  • Unified Field Perception and Application Analysis of Highway Traffic Risk Considering Risk Pulse Effects
    HU Li-wei, ZHOU Ze-yu, LIU Yi-chen, YANG Xiu-jian
    China Journal of Highway and Transport. 2026, 39(3): 116-134. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.009
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Autonomous Driving Trajectory Planning Based on Spatial Enhanced VLM
    JIANG Zheng-xin, NIU Ming-kui, HAN Pei-lun, GAO Bing-zhao
    China Journal of Highway and Transport. 2026, 39(3): 135-144. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.010
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Research on Path Planning Methods for Unpaved Road Based on Driving Risk
    WU Wen-guang, WANG Jia-kai, ZHANG Zhi-yong, HU Lin, ZHANG Jin-lai
    China Journal of Highway and Transport. 2026, 39(3): 145-160. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.011
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Path planning for autonomous vehicles on unpaved roads encounters significant challenges due to the presence of irregular surface potholes and bumps. To address this, a fusion path planning method that fully utilizes road surface features and considers their impact on driving safety is proposed. First, the unpaved road space was uniformly discretized in both the lateral and longitudinal directions, and an unpaved road risk field model was constructed based on an improved two-dimensional Gaussian distribution function to describe the risk distribution of road surface irregularities and road boundaries. Then, a safety cost calculation method that accounts for the vehicle's geometric dimensions was introduced. By combining path smoothness and lateral offset, a weighted multi-objective cost function was designed, and a dynamic programming algorithm was employed for path planning. Finally, incorporating actual road information, the Latin Hypercube Sampling method was used to generate 84 000 unpaved road models, each 30 meters in length with varying numbers and sizes of potholes and bumps, and path planning tests were conducted. Results demonstrate that the proposed method can effectively utilize the edges of potholes and bumps and traverse irregular features under severely degraded road conditions, resulting in paths with lower costs. Compared to the artificial potential field method and collision detection method, the average success rates are increased by 42.3% and 82.5%, while the average path lengths are increased by 27.9% and 57.6%. Consequently, the path planning capability of autonomous vehicles on unpaved roads is greatly improved, and the method exhibits superior adaptability and robustness.
  • Multi-vehicle Cooperative Decision-making in Mixed Traffic Environments via Hierarchical Game and Proximal Policy Optimization
    FANG Shi-yu, XU Cheng-kai, QIN Cheng, CUI Yi-ming, HANG Peng, SUN Jian
    China Journal of Highway and Transport. 2026, 39(3): 161-176. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.012
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    In open-road environments, autonomous vehicles (AVs) frequently encounter complex interaction conflicts and highly diverse human driver behaviors, leading to frequent decision failures and high takeover rates. Vehicle-to-vehicle (V2V) communication technologies offer the potential for cooperative driving, which can enhance the decision-making intelligence of AVs. To this end, this paper proposes an optimal cooperative decision-making framework based on hierarchical game representation and proximal policy optimization (PPO) to achieve multi-vehicle longitudinal acceleration planning in mixed human-autonomous traffic environments. The proposed method consists of two key modules: hierarchical game modeling for passing sequences representation, and reinforcement learning-based search for optimal passing sequences. First, considering the bounded rationality of human drivers, a hierarchical game framework is used to model the dependency among decisions, where reasoning depth is mapped to passing priority, enhancing the model's ability to accommodate heterogeneous human behaviors. Further, using PPO, the action space is defined over passing sequences, and merging, diverging, and crossing conflict scenarios are randomly generated during training to improve the method's adaptability to complex environments. Experimental results demonstrate that, compared to several baseline reinforcement learning methods, the proposed approach achieves faster convergence and higher rewards. Additional comparisons with representative cooperative decision-making algorithms show that our method offers superior performance in terms of both safety and efficiency. Finally, hardware-in-the-loop and driver-in-the-loop experiments validate the proposed algorithm's capability to handle real-world heterogeneous human drivers, highlighting its potential for practical deployment.
  • Resilient Arbitration and Safety Assurance of Driving Risks in Intelligent Vehicles via Secure Sandbox
    CAI Meng-chi, DENG Ning, YANG Dong-sheng, XU Qing, WANG Jian-qiang, LI Shen, LI Meng, LI Ke-qiang
    China Journal of Highway and Transport. 2026, 39(3): 177-193. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.013
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Research Progress on Traffic Regulation Compliance of Automated Driving Systems
    WANG Run-min, LIU Hui-min, CHENG Jing-jun, ZHU Yu, ZHAO Xuan, ZHAO Xiang-mo
    China Journal of Highway and Transport. 2026, 39(3): 194-213. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.014
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Research on Trustworthy Interaction Methods of Vehicle-road- cloud Data Based on Blockchain
    YUAN Xiao-fang, LI Zheng-yang, WANG Jin-lei, LI Zhe, FENG Ji, HAO Yu-xin
    China Journal of Highway and Transport. 2026, 39(3): 214-227. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.015
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To address the issues of low efficiency, repeated identity authentication, and untraceable malicious vehicles in existing vehicle-road interaction schemes, this paper proposes a blockchain-based trusted data interaction method for vehicle-road-cloud systems. First, a vehicle-road-cloud cyber-physical system (CPS) was constructed based on blockchain. This system integrates vehicles, roadside computing facilities (RCFs)-roadside units (RSUs), and certificate authorities (CAs). Registration was completed by backing up identity information required for verification on the blockchain. Second, a triple verification mechanism incorporating timeliness, pseudonyms, and hashing was proposed. This mechanism provides guarantees for vehicle-road authentication and interaction. To avoid repeated authentication, RCFs were introduced to assist in cross-domain authentication. Then, the vehicle selected random numbers. It combined pseudonym mechanisms and certificate systems to construct bilinear pairing and cryptographic operations. This process achieved trusted data interaction. Specifically, when RSU coverage is insufficient in remote areas, the system employs vehicle-to-vehicle (V2V) communication. A temporary vehicle identity generation mechanism was designed as a solution. This approach effectively ensures the security and reliability of connected data interaction without RSUs. Finally, simulations verified the proposed scheme. Experimental results demonstrate that the scheme meets the security requirements of the vehicle-road-cloud CPS interaction process. Compared to existing schemes, it reduces time overhead by 47.76% and space overhead by 21.69%.
  • Hierarchical Distributed Model Predictive Energy-saving Control for Connected Vehicles
    BIAN You-gang, DENG Xiao-yang, TAN Yan, WEN Shu-ting, LIU Qun-xin, CHEN Chao-yi
    China Journal of Highway and Transport. 2026, 39(3): 228-240. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.016
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    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.
  • Decoupled Estimation of Commercial Vehicle Mass and Slope Using a Two-layer Structure
    JIN Yan-lin, LI Yi-nong, ZHENG Ling, HE Bo-hao, YANG Xian-tong, LI Guang-xuan
    China Journal of Highway and Transport. 2026, 39(3): 241-259. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.017
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Accurate acquisition of vehicle mass and road slope is crucial for optimizing the dynamic control of intelligent connected vehicles and improving energy efficiency. However, for heavy commercial vehicles, the strong coupling between vehicle mass and road slope, the wide range of mass variation, and the insufficient robustness of algorithms in real-world environments pose significant challenges to the precise estimation of mass and slope. To address this, a two-layer decoupled estimation method for mass and slope was proposed, which achieved decoupling and precise estimation through hierarchical data- and model-driven approaches. The first layer estimated vehicle mass using a model trained on real vehicle data. To handle the issue of out-of-distribution (OOD) input data, K-means clustering and state-space shortest distance calculation were employed to obtain a confidence metric. Based on this, a τ-Bidirectional Long Short-term Memory (τ-BiLSTM) model with integrated confidence was constructed. The second layer estimated road slope based on the mass estimation results from the first layer. A slope estimation algorithm, based on the vehicle longitudinal dynamics model, was developed using a Square Root Cubature Kalman Filter (SCKF). Furthermore, the slope estimation results derived from kinematics were fused into the SCKF framework as an initial value. Finally, the algorithm was validated using real vehicle data collected in real-world environments. The results show that the proposed mass estimation method can effectively suppress estimation errors caused by OOD input data. On a self-built test set, 94.11% of the sample points have estimation errors within 10%, and the RMSE of mass estimation under typical operating conditions is less than 0.26 t. For slope estimation, the RMSE under typical operating conditions is less than 0.39%. The research results verify the high accuracy and robustness of the proposed method in complex dynamic environments, providing reliable dynamic parameter support for intelligent dynamic control and energy optimization management of heavy commercial vehicles.
  • Vehicle Trajectory Reconstruction Method Driven by Multi-source Sparse Data and Matrix Decomposition
    YIN Ju-yuan, YANG Hong-zhi, LI Bing, LI Bo-yang, CHEN Yu-qian
    China Journal of Highway and Transport. 2026, 39(3): 260-275. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.018
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Full-sample vehicle trajectories are fundamental information for accurately depicting the overall and detailed operation of traffic flow. To address the challenges of acquiring full-sample vehicle trajectories in the current multi-source sparse data environment, this study proposes a hybrid-driven approach combining car-following models and matrix decomposition. The method reconstructs complete vehicle trajectories by capturing the intrinsic relationship between microscopic traffic flow characteristics and multi-source sparse data. First, spatiotemporal alignment is performed on the multi-source heterogeneous sparse data provided by fixed and mobile sensors to establish a sparse trajectory matrix data structure. Then, the intelligent driver model generates missing vehicle trajectory data in blind zones between upstream and downstream fixed detectors, filling the sparse trajectory matrix to create an initial trajectory matrix. Next, a truncated singular value decomposition algorithm is employed to iteratively optimize the initial trajectory matrix, resulting in full-sample vehicle trajectories. Finally, this method is validated in both the real-world data environment and the simulation environment. The results show that in the real-world data environment, when the distance between fixed sensors is 250 m and the penetration rate of connected vehicles (CV) is 15%, the full-sample vehicle trajectories reconstructed by this method have MAE, MAPE, and RMSE of 4.40 m, 2.91%, and 6.17 m, respectively. Compared to existing advanced reconstruction method, these metrics are optimized by 37.68%, 9.35%, and 48.06%, and the reconstructed trajectories under congested flow are more accurate and smoother. In the extreme sparse data environment of simulation, with two fixed sensors distance of 600 m and CV penetration rate of 5%, this method still maintains high reconstruction accuracy, achieving the three metrics of 10.73 m, 1.92%, and 15.98 m, respectively. The research findings can provide robust support for precise perception and fine-grained management of road traffic.
  • A Cloud-based Path Tracking Control Method for Intelligent Connected Vehicles Considering Full-link Vehicle-to-cloud Latency
    ZHANG Xin-rui, XIONG Lu, ZHANG Pei-zhi, WANG Xiu-rong, TIAN Meng-jie, YIN Dong-xiao, XIONG Hao
    China Journal of Highway and Transport. 2026, 39(3): 276-288. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.019
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Cloud-based path tracking control systems can leverage the rich information and powerful computing capabilities of vehicle-cloud integrated architectures to effectively enhance the path tracking performance of intelligent connected vehicles. However, dynamic delays inherently existing in vehicle-cloud communication links can severely degrade control stability and tracking accuracy. To address the influence of dynamic delays over the entire vehicle-cloud communication link, a cloud-based path tracking control method based on predictive compensation and adaptive optimization was proposed. First, based on large-scale real vehicle test data, a gamma distribution model of communication delay and a first-order response model of vehicle actuator delay were established. Then, a Bayesian long short-term memory network with attention mechanism was designed, which incorporated network signal strength indicators to enhance delay prediction accuracy and robustness. Subsequently, the stability criterion of the path tracking system under delay disturbance was analyzed, and an adaptive optimization strategy of linear quadratic regulator (LQR) control weights based on predicted delay was proposed to achieve a dynamic balance between tracking accuracy and system stability. Finally, simulations and real vehicle tests were carried out for validation. Simulation results show that, compared with the traditional fixed-weight LQR method, the proposed method reduces the lateral displacement RMSE by 63% while maintaining the heading error within 0.03 radians under normal small-delay conditions. Under abnormal large-delay conditions, the lateral root mean square error (RMSE) and maximum peak error (MPE) are reduced by 46% and 61%, respectively, while the heading error RMSE and MPE are reduced by 76% and 71%, respectively. The real-world experimental results demonstrate that the proposed method outperforms the fixed-parameter LQR controller in two representative scenarios-U-turn and double lane change-under various network conditions. The lateral error RMSE and MPE are reduced by up to 69% and 52%, respectively, while the heading error RMSE and MPE are reduced by up to 40% and 33%, respectively. These results provide strong evidence that the proposed cloud-based path tracking control approach effectively adapts to the dynamic characteristics of vehicle-cloud delays, significantly enhancing the stability and accuracy of cloud-based path tracking.
  • A Hierarchical Modeling and Control Strategy for Vehicle-road-cloud Integration CPS on Highways
    HE Chang-chang, ZHAO Min, SUN Di-hua, HU Chuang, CHEN Zhang-shun
    China Journal of Highway and Transport. 2026, 39(3): 289-301. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.020
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To address highway congestion caused by traffic-flow bottlenecks and demand imbalance, and to meet real-time control requirements under limited communication and computation resources, this study proposed a hierarchical modeling and dynamic centralized-distributed control strategy for vehicle-road-cloud integration cyber-physical systems. The strategy constructed a three-layer collaborative architecture comprising the system, subsystem, and vehicle levels. At the system level, a traffic-state-based dynamic clustering mechanism was used to balance traffic efficiency with communication and computation costs; at the subsystem level, variable speed limits and ramp metering were jointly optimized to reduce total travel time and increase total travel distance, while a soft constraint suppressed ramp queues; at the vehicle level, the car-following controller considered both the preceding-vehicle state and the optimal speed limits from the upper layer, improving the implementability of speed advisories in an intelligent connected vehicle environment. In a 25 km highway simulation scenario with a typical bottleneck, the proposed strategy was compared with no control, local Model Predictive Control (MPC), Feasible Cooperation-Based Model Predictive Control (FC-MPC), and a global Proximal Policy Optimization (PPO) controller. The results show that, relative to no control, the proposed strategy reduces total travel time by up to 12.34% and increases total travel distance by up to 0.9%, and relative to the global PPO controller, it reduces the number of control updates by up to 93.33%. Overall, the proposed strategy outperforms no control and local MPC in traffic-efficiency indices, achieves performance comparable to FC-MPC and PPO, and incurs lower communication and computation burdens with a tunable trade-off via weight adjustment; simulations under communication interruption and perception delay further confirm its robustness. This study shows that the strategy effectively handles bottleneck-dominated and high-demand scenarios under limited communication and computation resources, with good engineering applicability and robustness.
    • Pavement Engineering
  • Working Performance Evaluation and Multidimensional Index Correlation Analysis of Full-series Resin-modified Emulsified Asphalt
    WANG Chao-hui, WANG Yu-cong, ZHANG Xin-yong, ZHANG Shao-bo, CHEN Qian, FU Hao
    China Journal of Highway and Transport. 2026, 39(3): 302-317. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.021
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To comprehensively evaluate and accurately determine the performance of resin-based modified emulsified asphalt, and to further enhance the efficiency of demand-driven development and the service quality of high-grade resin materials, a full series of resin-based modified emulsified asphalts was systematically developed. These included formulations with single, binary, and ternary waterborne polymers. The mechanical, thermosensitive, waterproofing, rheological, and fatigue properties of the emulsified asphalts were examined. The effects of different waterborne resin polymer modifiers on cohesive strength, deformability, toughness, elasticity, low-temperature impact resistance, adhesion, and bonding were clarified. The temperature sensitivity, waterproofing ability, rheology, and fatigue life of resin-based modified emulsified asphalts were comprehensively evaluated. Based on the Pearson correlation coefficient method, the key indicators of mechanical and rheological properties were identified. A multidimensional performance correlation system was established, and functional relationships among critical performance indicators were revealed. The results show that different types of resin-based modified emulsified asphalts exhibit significant performance differences. Waterborne epoxy resin (WER) improves the mechanical, waterproofing, and high-temperature properties of emulsified asphalt. In particular, the binary WER/waterborne polyurethane (WPU) modified emulsified asphalt achieves a tensile strength of 5.98 MPa and a pull-off strength of 1.95 MPa at 25 ℃, with a creep recovery rate (R) of 97.36% at 76 ℃ and 3.2 kPa. The WPU-modified emulsified asphalt series exhibit superior cohesive strength, elasticity, toughness, low-temperature crack resistance, and fatigue performance, achieving an impact strength of 12.50 kJ·m-2 at -20 ℃. Upon incorporating waterborne acrylic (WA) or styrene-butadiene rubber (SBR), the WER/WPU/WA and WER/WPU/SBR ternary composite-modified emulsified asphalts demonstrate excellent flexibility, withstanding 180° bending without fracture at temperatures as low as -20 ℃. Additionally, their stiffness modulus (S) at -24 ℃ remained below 300 MPa, while the creep rate (m) exceeded 0.3. The WER/WPU/WA-modified emulsified asphalt exhibits outstanding toughness and fatigue life. Notably, quadratic functional relationships are observed between binder detachment rate and waterproofing performance, tensile strength and phase angle (δ) at 70 ℃, as well as impact strength at -20 ℃ and S at -18 ℃, with coefficients of determination (R2) of 0.97, 0.94, and 0.93, respectively. A power-law relationship is identified between fatigue life and complex shear modulus (G*) at 70 ℃, with R2 of 0.89.
  • High-precision and Real-time Intelligent Pavement Condition Survey
    ZHANG Ao-nan, DONG Zi-shuo, ZHANG Wen-jin, SHANG Jing, ZHAN You, AI Chang-fa, HUI Bing
    China Journal of Highway and Transport. 2026, 39(3): 318-332. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.022
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    The advancement of intelligent pavement condition survey faces a core challenge in achieving coordinated improvements in accuracy, functionality, and timeliness. To address this, a real-time intelligent pavement condition survey framework was established based on a 3D laser imaging system, and achieved key breakthroughs in high-precision 3D laser imaging, massive 3D data compression and storage, and real-time intelligent recognition algorithms for multi-target identification including pavement distresses and surface design features. The 3D laser imaging system was capable of acquiring high-precision 2D and 3D pavement data in real time at a speed of 120 km·h-1, which provided robust data support for intelligent recognition and refined assessment. While ensuring data quality, the massive 3D data compression and storage algorithm significantly reduced storage and transmission demands, enhanced memory efficiency, and ensured stable data acquisition during continuous high-speed operation. The proposed multi-target real-time intelligent recognition algorithm achieved pixel-level accuracy in identifying various pavement objects including pavement background, crack, pothole, sealed crack, patch, scratch, marking, manhole, expansion joints, and concrete slab joints, with an average F1-score of 94.34%. After model pruning and quantization, the model achieved an inference speed of 6 ms per frame. Field verification demonstrated that the integrated processing speed of the proposed high-precision and real-time intelligent pavement condition survey system reached 160 km·h-1, and that it was capable of completing full-process operations including data collection, intelligent analysis, and report generation for 400 km of road pavement within a single day. The results show that the framework meets both high-precision and high-efficiency requirements, and exhibits strong robustness in recognizing complex pavement distresses. The research provides an efficient and feasible technical reference for comprehensive, accurate, and timely assessment of pavement conditions, thereby supporting preventive maintenance strategies for pavement management systems.
    • Bridge Engineering
  • Seismic Retrofit of Highway Bridges: State of the Art Review
    SU Jun-sheng, LOU Ce-xiang, HAN Qiang, XIANG Nai-liang, LI Zhen-xin, LI Zhong-xian
    China Journal of Highway and Transport. 2026, 39(3): 333-357. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.023
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Bridge structures are critical components of lifeline engineering, and their seismic performance is vital to the safety and resilience of transportation networks and lifeline systems. Approximately 60% of China's territory lies in high seismic intensity zones, and in-service highway bridges are therefore exposed to considerable seismic risks. By the end of 2023, there were 1.079 3 million in-service highway bridges nationwide, among which a large number were either not designed with seismic considerations or were designed with insufficient seismic fortification levels, making their seismic safety difficult to guarantee. Consequently, the demand for seismic retrofitting of in-service bridges is significant. To address this issue, this paper first reviews the historical development of seismic retrofitting practices and related codes both in China and abroad. It then summarizes commonly adopted seismic retrofitting techniques for bridges, including unseating preventing, pier strengthening, cap beam and beam-column joint strengthening, foundation enhancement, as well as the application of isolation and energy dissipation devices. On this basis, the paper synthesizes recent research progress in seismic retrofitting of bridges, covering the development of retrofitting theories, advancements in strengthening of substructures, seismic control and replaceable devices, retrofitting technologies for arch bridges and long-span flexible bridges, as well as the application of high-performance materials and rocking self-centering technologies. Finally, the paper identifies existing challenges in China's seismic retrofitting research for bridges, including the absence of systematic codes and the lack of a comprehensive design framework. Future directions are proposed, such as establishing unified codes, developing an integrated retrofitting design system, advancing resilience-based retrofitting methodologies, and incorporating the concept of life-cycle service, with the ultimate goal of ensuring the seismic safety of in-service bridges.
  • Research on Cross-bracing Design Method of Curved Composite Steel I-girder Bridges
    FENG Bo-wen, LIU Yong-jian, CUI Xiao-xiao, ZHANG Bin-bin
    China Journal of Highway and Transport. 2026, 39(3): 358-369. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.024
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    During the concrete deck construction phase, the curved steel I-girders carry the first-stage dead load. The cross-bracing transmits torque by limiting the relative rotation (cross-section distortion) between the steel girders, serving as the primary load-bearing component of the structure. To determine the optimal spacing for the cross-bracing, this paper first established a mechanical analysis model for the curved steel girders under dead load. The theoretical analytical method was derived based on Vlasov's thin-walled beam theory, and the results were validated using a refined FEA model. Then, based on the characteristics of the sectional normal-stress distribution, an indicator called the “distortion stress ratio” was proposed to evaluate the relative magnitude of distortion effects compared to bending-torsion effects. The influence of structural layout and design parameters on this indicator was also investigated. Finally, a simplified design method for the spacing of cross-bracing was obtained through parameter fitting and compared with existing design methods. The results show that under dead load, the distributions of bending normal stress, non-uniform torsional warping normal stress, and distortional warping normal stress in the flange of the steel I-girder exhibits a combination of average normal stress and transverse bending normal stress. Adding cross-beams can effectively constrain the cross-section distortion, reduce distortional warping normal stress, and change the distribution of transverse bending normal stress. When designing the strength of the steel I-girder, the maximum value among the three types of stresses mentioned above should be used for calculation.The distortion stress ratio decreases exponentially with the increase in the number of transverse connections, curve radius, and flange width, while it increases linearly with the increase in the height of the steel girder. Given the dimensions of the steel I-girder cross-section and the curve radius, the spacing of the cross-bracing for the curved steel-concrete composite girder bridge can be determined using the formula presented in this paper by limiting the maximum distortion stress ratio. The method proposed in this study has an error of less than 5% compared to the design values and demonstrates higher accuracy than methods in AASHTO 2017 and the JASBC manual.
  • Research on Axial Tensile Cracking Behavior and Calculation Method of Reinforced Ultra High Performance Concrete
    WEI Chuan, ZHANG Qing-hua, GENG Bo, CUI Chuang, CHENG Zhen-yu
    China Journal of Highway and Transport. 2026, 39(3): 370-382. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.025
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To explore the differences in the axial tension cracking behavior of reinforced Ultra High Performance Concrete (R-UHPC) components with different fiber contents and improve the crack width calculation method for R-UHPC with low fiber content, axial tension tests were carried out on UHPC materials and R-UHPC components with different fiber contents. The macroscopic mechanical behavior and cracking behavior of R-UHPC components before yielding were observed. The axial tension cracking characteristics and mechanisms of each group of components were analyzed. Based on the bond-slip theory, the calculation of cracking load, crack spacing, and strain difference, as well as the selection of fiber bridging stress and average bond stress in the calculation of the maximum crack width of R-UHPC components with low fiber content were discussed. The research results indicate that: the cracking process of R-UHPC components with low fiber content can be divided into a single-crack stage and a multi-crack stage. Their cracking behavior is jointly controlled by the bond between the steel rebar and UHPC and the fiber bridging effect. The average crack width is in the range of 0.02-0.11 mm, and the maximum crack width is in the range of 0.03-0.15 mm. Adding 1%-2% volume fraction of steel fibers to R-UHPC is sufficient to control the maximum crack width within 0.1 mm. The cracking process of R-UHPC components with high fiber content only has a multi-crack stage, and their cracking behavior is only controlled by the fiber bridging effect. The average crack width is not more than 0.02 mm, and the maximum crack width is not more than 0.03 mm. R-UHPC components with high fiber content do not need to check the crack width under the serviceability limit state. The shrinkage effect will significantly reduce the cracking load of R-UHPC components. Under the assumption that the fiber bridging stress at the crack section is constant, the calculation formulas for the maximum crack width in the single-crack stage and the multi-crack stage remain the same in form, but the change in the average bond stress between the steel rebar and UHPC needs to be considered to reflect the impact of multi-crack formation on the crack spacing.
  • Bonding Performance and Shear Mechanism Study of Salt Freeze-thaw Damaged Concrete Interface Reinforced by ECC
    HE Lan-qing, LEI Hai-peng, WU Fang-wen, LI Qiang, LI Zi-run, CHEN Ao
    China Journal of Highway and Transport. 2026, 39(3): 383-397. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.026
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To investigate the interfacial shear performance of salt-freeze-thaw-damaged normal concrete (SftdNC) reinforced with engineered cementitious composite (ECC), an experimental study was conducted on 66 ECC-SftdNC shear specimens. The effects of salt freeze-thaw cycles, interfacial roughness, ECC compressive strength grade, interfacial agent type, and ECC curing age on both macroscopic shear performance and microscopic interfacial adhesion were analyzed. The test results indicate that the failure modes of the ECC-SftdNC interface can be categorized into material failure, interface failure, and hybrid failure. Scanning electron microscopy (SEM) analysis revealed that the microscopic bonding mechanism at the ECC-SftdNC interface primarily depends on the combined action of both chemical bonding action and mechanical interlock action. The application of interfacial agents can effectively fill interfacial cracks and enhance the interfical transition zone. After 200 salt freeze-thaw cycles, the ECC-SftdNC interfacial shear strength and shear fracture energy decrease by 68% and 28%, respectively. Increasing the interfacial roughness from 1.5 mm to 6.5 mm enhances the interfacial shear strength by 92%. Among different interfacial agents, fly ash-modified cement paste demonstrates the most significant improvement in interfacial shear performance, increasing shear strength and fracture energy by 54% and 75%, respectively. Based on plastic limit theory, a calculation method for ECC-SftdNC interfacial shear strength was established, which showed good agreement with experimental values. The multi-parameter influence mechanisms and theoretical model proposed in this study provide crucial theoretical foundations and technical support for the ECC strengthening design and engineering applications of concrete bridges subjected to salt freeze-thaw damage in cold regions.
  • Experiment of Fire Resistance and Residual Load-carrying Capacity of PC Composite Box Girder with Corrugated Steel Webs
    ZHENG Shang-min, LIU Cui-chen, GUAN Chong, WU Zhi-qiang, ZHAO Xian-qi, LIU Wei-liang, CHENG Hai-gen
    China Journal of Highway and Transport. 2026, 39(3): 398-411. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.027
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To study the fire resistance performance of a PC composite box girder with corrugated steel webs under fire throughout the process and its post-fire residual bearing capacity, a simply supported PC composite box girder with corrugated steel webs was designed and fabricated. The girder underwent hydrocarbon fire resistance testing followed by post-fire static loading experiments. The temperature distribution across the cross-section, mid-span deflection evolution, and effective prestress decay behavior of the test girder under hydrocarbon fire were analyzed. The residual load-bearing capacity and complete flexural failure process after the fire were also explored. The fire resistance test results showed a cross-sectional temperature gradient along the height of the test girder during the fire and the post-fire cooling process. The temperature of the corrugated steel webs increased significantly faster than that of the concrete top and bottom plates when exposed to fire. The maximum cross-sectional temperature gradient under fire was 626 ℃. Within 1 h after the fire was stopped, the cross-sectional temperature gradient decreased. When the test girder was subjected to fire for 70 min under a load ratio of 0.195, the mid-span deflection was about 26 mm. Post-fire, the load was maintained for 1 h, and the girder's deflection continued to increase to approximately 44 mm. After removing the load, the girder rebounded at 70.5%, and the final residual deformation was about 13 mm. The effective prestress of the prestressing strands under fire increased slightly and decreased rapidly. Post-fire was extinguished, the prestress decreased to about 406 MPa and remained stable, with a total decrease of 60.3%. The post-fire static load test results showed that the entire process of the test girder's flexural failure under post-fire external load can be roughly divided into four stages: a quasi-elastic stage, a crack propagation stage, a local nonlinear stage, and a failure stage. The remaining bearing capacity was about 60.0% of the ultimate bearing capacity at room temperature. The stress change process of the prestressing strands after the fire can be divided into a stage of coordinated stress, a stage of rapid redistribution of stress, and a stage of the main bearing of the prestressing strands. When the girder body failed, the concrete of the bottom plate was pulled apart, but the prestressing strands had not yet broken, and its stress was 1 345.7 MPa, which is about 72.3% of the standard value of tensile strength of 1 860 MPa. There is a close relationship between the rate of change of the stress of the prestressing strands and the degree of damage to the girder. This study provides a scientific reference for assessing the remaining bearing capacity and bridge repair of post-fire composite box girders with corrugated steel webs.
    • Tunnel Engineering
  • Refined and Advanced Prediction Method of Tunnel Surrounding Rock Classification Based on Drilling Parameters
    WANG Ming-nian, ZHONG Hao, SHEN Hong-lin, HUANG Bing-xu, YI Wen-hao, XIA Qin-yong, TONG Jian-jun
    China Journal of Highway and Transport. 2026, 39(3): 412-423. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.028
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    To continuously forecast the geological condition of the next unexcavated cycle in the tunnel construction stage and considering the non-uniformity of a large section of surrounding rock, a refined and advanced prediction method for tunnel surrounding rock classification based on drilling parameters is proposed. Based on two drilling and blasting tunnels in southwest China, 2 670 cycles of drilling parameters and their surrounding rock classifications were collected, and the statistical features of the collected drilling parameters were extracted. An advanced prediction model for tunnel surrounding rock classification based on drilling parameters was constructed using long short-term memory in deep learning. The model can be used to predict the surrounding rock classification of the next unexcavated cycle using cycle sequence data composed of the statistical characteristics of the drilling parameters of four excavated cycles and one drilled cycle without blasting as the input of the model. Based on this, using the three-dimensional spatial characteristics of drilling parameters, a three-dimensional spatial unit partition method for the surrounding rock is proposed. The entire cycle is divided into several sections in the horizontal direction, and a refined advanced prediction strategy for surrounding rock classification is proposed. Finally, using the trained advanced prediction model of the tunnel surrounding rock classification based on the drilling parameters, a refined advanced prediction of the next unexcavated cycle surrounding rock classification was realized. The results show that the LSTM model is effective in predicting the surrounding rock classification in advance and that refined and advanced prediction method for tunnel surrounding rock classification based on drilling parameters can reveal the geological conditions and heterogeneity characteristics of the unexcavated cycle in advance.
  • Automated Segment Segmentation and Deformation Extraction Algorithms for Shield Tunnel Point Clouds Based on Customised Label Encoding and RandLA-Net
    LIN Wei, XIE Xiong-yao, GUAN Zhen-chang, CHANG Jia-qi
    China Journal of Highway and Transport. 2026, 39(3): 424-437. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.029
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Shield tunnels are an important component of highway infrastructures and their structural conditions are fundamental to the functioning of regional traffic and urban utility. 3D laser scanning presents a fast and accurate technology to acquire massive 3D spatial data of structural surfaces, whereas the heavy burden of processing point clouds hinders its scalable application. An end-to-end segmentation method for shield tunnel point clouds is proposed based on 3D computer vision and advanced RandLA-Net network architecture to enhance the intelligence of deformation measurement for shield tunnels. A customised label encoding method and a corresponding loss function tailored to the nature of shield tunnel point clouds are proposed to improve the efficacy of deep learning model training. Based on the proposed segmentation method for shield tunnel point clouds, this paper further proposes an algorithm for automatically extracting segment deformation to achieve the comprehensive measurement of structural conditions. Data experiments were conducted based on 443 rings of real-world point clouds, which were manually annotated, to validate efficacy and investigate influencing factors of the proposed method. The results indicate that trained deep learning models based on RandLA-Net can effectively realise the segment segmentation of shield tunnel point clouds. The customised label encoding can improve the segmentation efficacy of the trained deep learning models, with notably improved accuracy near segment edges. The proposed deformation extraction algorithm achieved fine-scale and accurate segment-wise deformation measurement, providing abundant engineering information such as segment rotation, segment deflection and joint dislocation. This paper provides technical means for dynamic construction management, digital as-built archiving, and fine-scale operational management for shield tunnels, contributing to the enhancement of digitisation and intelligence in highway infrastructures.
  • Research on the Mechanical Performance of Joints in Fully Precast Rectangular Prefabricated Subway Station
    LING Tong-hua, ZHOU Zhi-hui, TAN Jia-nuo, XIE Zi-long, LIAO Yi-xuan
    China Journal of Highway and Transport. 2026, 39(3): 438-448. https://doi.org/10.19721/j.cnki.1001-7372.2026.03.030
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    The fully prefabricated rectangular assembled metro station, as a new structural form, effectively improves construction quality and speed. However, the influence of its joint connections remains unclear. To explore the optimal joint type and its mechanical performance, full-scale indoor tests were conducted to comparatively analyze the mechanical deformation characteristics of joint connections under six types of connections (two types of mortise-tenon connections and three types of bolted connections). Based on the engineering background of a rectangular assembled metro station in Shenzhen, numerical simulations were employed to analyze the station structure's responses to different joint connection types. The results indicate that: ① The joint components of all six connection types meet the bearing capacity requirements under normal operational conditions of the metro station structure. The single mortise-tenon joint exhibited slightly lower deformation compared to the double mortise-tenon joint during the linear bearing stage. Bolts effectively limited the mid-span deflection of joint components, with curved bolts showing the best suppression effect. ② During the ultimate bearing stage, curved bolts provided the most significant enhancement to the joint components' ultimate bending moment capacity, with an increase of 23.9%. Under the same bolt conditions, the ultimate bending moment capacity of single mortise-tenon joints was higher than that of double mortise-tenon joints. ③ The bending stiffness curve of the joint components consists of three stages, characterized by significant variable stiffness. The bending stiffness reaches its peak when the bolted components enter the rapid decline phase. The single mortise-tenon joint bolted component outperforms other conditions in terms of ultimate bending moment capacity, bending stiffness, and deformation capacity. ④ In fully prefabricated rectangular assembled metro stations connected by six types of joint components, the maximum principal stress values were primarily concentrated at the corner of the roof slab, on both sides of the middle slab, and at the bottom side of the base slab. The minimum principal stress values were predominantly located at the inner corner of the roof slab, the outer edge of the roof slab, and the connection between the base slab and the middle column. The overall station structure exhibited optimal stress performance under the configuration of curved bolt single mortise-tenon joints.
News More>>
Download More>>
Links More>>
New Media
  • Wechat Account
  • WeChat video Account
Visited
    Total Visits:    Today:    Online:
Web Copyright © China Journal of Highway and Transport, All Rights Reserved.
Address: Chang'an University, Southern Middle Section of the Second Circle Road, Xi'an, P.R.China Zip Code: 710064
Tel: 029-82334387　
E-mail：zgglxb@163.com；zgglxb@qq.com
Powered by Beijing Magtech Co. Ltd,.