IET Intelligent Transport Systems最新文献

Heavy-haul trains play a crucial role in long-distance bulk transportation, yet their enormous mass and kilometer-scale length lead to complex longitudinal interactions and high coupler forces, which threaten operational safety. Conventional mechanism-based models, while accurate, are computationally expensive and unsuitable for real-time prediction. To address this limitation, this study develops a data-driven prediction framework that combines physics-based modelling and deep learning. A detailed longitudinal dynamics model of a 20,000-ton train operating on the Shuohuang Railway is constructed, incorporating traction, electrical braking, and resistance characteristics to compute coupler forces under varying gradients and curvature conditions. Based on this model, a QP-based optimization algorithm and a high-fidelity simulation platform are used to generate multi-strategy operating datasets that balance energy efficiency, punctuality, and ride comfort. The resulting data are processed using normalization and sliding-window segmentation to form supervised learning samples. A multi-resolution dual-stream LSTM (MRDS-LSTM) and its attention-enhanced variant (MRDS-LSTM–Attn) are then proposed to capture both short-term fluctuations and long-term temporal trends. Compared with RNN, GRU, LSTM, Bi-LSTM, NLSTM, CNN-LSTM, CNN-NLSTM, CapNet-NLSTM, Transformer, and Informer baselines, the proposed model achieves the highest prediction accuracy with MRDS-LSTM-Attn achieves an MAPE of 2.57%, and $�R^2$ of 0.9888. The results demonstrate that the proposed framework effectively bridges physical modelling and data-driven prediction, achieving up to 706$�\times$ faster inference than traditional solvers. It provides a practical foundation for intelligent heavy-haul train operation, supporting real-time coupler force monitoring, predictive safety control, and future extensions to pneumatic braking and field data validation.

To enhance the service scope and quality of urban public transport systems, this study investigates the optimal design problem of feeder-bus networks related to urban rail transit considering time windows (FBNDP-TW). To ensure an acceptable passenger travel time, we differentially set the travel time window for each origin-destination (OD) pair based on the ideal travel time. Considering logical constraints, capacity constraints and time window constraints, we construct an FBNDP-TW optimisation model to minimise passengers’ generalised travel cost and bus operators’ operating cost. To solve this model, a genetic algorithm is developed with a diverse multi-neighbourhood crossover operation that includes ‘direct’, ‘forward’ and ‘adjacent’ rules. This crossover operation mechanism can efficiently make the feeder-bus network quickly meet time window constraints to guarantee its quality. Finally, the proposed model and algorithm are evaluated using a standard example network. The results confirm that they can effectively ensure the travel time of each OD. Although integrating time window constraints slightly raises network cost, it significantly reduces the maximum OD detour ratio and ensures the travel time of all ODs within the acceptable range.

To address the issues of false positives and missed detections of multi-scale cracks and small targets in complex environments, this paper proposes an enhanced YOLOv10 instance segmentation network named YOLO-RCS (YOLOv10s road crack segmentation), specifically designed for segmenting surface cracks. YOLO-RCS utilizes the DCNv4 module to enhance feature extraction in the backbone network, improving the accurate localization of surface crack segmentation. Additionally, we introduce a novel C3FB structure (an efficient fusion of the C3 module and FocalNextBlock structure) to replace the C2f module in YOLOv10's neck network, aiming to reduce the number of parameters while enhancing model accuracy. Finally, we improve the original loss function to the WIOU loss function, which increases the model's precision and mean average precision (mAP) for segmenting surface cracks. Experimental results show that our model achieves an mAP50 of 90.0% on the surface crack segmentation dataset Crackseg9k, a 5.0% improvement over the original algorithm, with a precision of 91.6%, demonstrating excellent segmentation performance. Compared to some mainstream object detection algorithms, our proposed method also exhibits certain advantages.

Computer-based systems (CBSs) are complex and critical, with risks to human lives and the environment. Ensuring their safety requires rigorous methods. Unlike traditional approaches that model system mission specifications in a single step before introducing safety requirements, this paper proposes a layered strategy for modelling both mission and safety requirements. This strategy ensures alignment between safety and mission requirements at each layer and formally proves their sufficiency with respect to system-level safety constraints (SLSCs), thereby achieving synchronised assurance of functionality and safety. Mission requirements are specified in Event-B, while System-Theoretic Process Analysis (STPA) derives safety requirements (SRs) to address system-level hazards. These SRs and SLSCs are integrated into the Event-B model to ensure consistency and verify compliance. By iteratively applying this pattern at each STAMP refinement step, a layered CBS is developed with safety as a core feature. Key contributions include stepwise STAMP refinement aligned with the system architecture hierarchy, coordinated development of Event-B models and STPA analysis using common STAMP models, and managing abstraction levels to ensure compliance between SRs and SLSCs while addressing formal verification complexity. A case study of a computer-based interlocking system demonstrates the approach's practical application.

Low-light road segmentation is a challenging dense prediction task, which is very important for the safety monitoring of intelligent sea ports at night and the autonomous vehicles of shipping logistics. Most current research focuses on scenes with sufficient light. At the same time, there are few datasets for low-light scenes, making research on low-light perception very difficult and seriously restricting the shipping logistics industry's ability to operate safely at night. Directly applying segmentation methods developed for well-lit scenes to low-light road segmentation is unsatisfactory. To solve this problem, we propose a new approach by building an image enhancement module and an edge detection module, and integrating them into existing well-lit segmentation models as a plugin to meet the road segmentation requirements for low-light scenes. Specifically, to compensate for the lack of low-light image detail, we design an image enhancement module that achieves end-to-end pixel-level image enhancement by connecting four image processing filters in series and using convolutional neural network to predict hyperparameters. Additionally, to address the problem that road edges become blurred and difficult to extract in low-light images, we design an edge detection module to maximize its ability to extract road edges by selecting differential pixel pairs using different strategies and efficient combinations. We conduct comprehensive experiments on our newly released dataset, LoRD, demonstrating that our method significantly outperforms previous state-of-the-art models with relatively few parameters and computational cost. Our method achieves new SOTA performance in terms of accuracy and computational efficiency, achieving 93.29$�\%$ at 78.44 FPS in DDRNet-23-slim. The source code and dataset will be publicly available.

This paper introduces a robust Fuzzy Sliding-Mode Controller (FSMC) optimised through a Genetic Algorithm (GA) for a nonlinear full-vehicle semi-active suspension system, explicitly accounting for real-world uncertainties. Specifically, the study considers three types of uncertainties, including variations in sprung mass, temperature variations, and vehicle obsolescence. The proposed control method has been combined with two double-input-single-output fuzzy controllers to enhance both ride comfort and road-holding performance. A seven-degree-of-freedom nonlinear full vehicle model, including a semi-active suspension system with a nonlinear spring and linear magnetorheological (MR) damper, is presented. The study aims to investigate the impact of these uncertainties on the semi-active suspension system. A key innovation stems from the GA-based tuning of FSMC parameters, which dynamically adapts the controller's behaviour for optimal performance under varying uncertain conditions. The findings indicate noteworthy enhancements, such as a roughly 7% increase in ride comfort for FSMC 2 compared to FSMC 1 and an even more substantial, up to 12% improvement in road-holding performance in State 1. Similarly, in State 2, FSMC 2 achieved a 10% improvement in ride comfort and up to a 12% enhancement in road-holding performance over FSMC 1.

To solve the problem of lane changing cooperative control of autonomous vehicles in high-density heterogeneous traffic flow, by analyzing the characteristics of the mandatory lane changing behavior of autonomous vehicles, a dual-lane utility calculation model based on driving style was established, and a lane changing cooperative control game strategy was proposed. Through joint simulation experiments using VISSIM and MATLAB, the results indicate that, in mixed driving environments, the driving style-based game model significantly enhances lane changing performance compared to the traditional MOBIL model and ordinary game models. On average, the lane changing position is advanced by approximately 100 m, and the delay is reduced by 4 s. Meanwhile, safety distance thresholds of 80 m and 50 m were set for aggressive and conservative drivers, respectively, effectively balancing safety and efficiency. Furthermore, by analyzing the interactive effects between the initial position of lane changing vehicles and driver styles, it was found that aggressive drivers need to abandon lane changing when their initial position is within the range of [0, 80] m, while conservative drivers can ensure safety even when their initial position is within [0, 50] m.

Optimizing regenerative braking in dual-motor electric vehicles (EVs) is critical for extending driving range but presents a complex high-speed control problem. This study proposes a novel, real-time control strategy by training a hybrid deep neural network–decision tree (DNN–DT) model on an optimal dataset generated by offline dynamic programming (DP) considering seven key characteristic variables: road grade, friction coefficient, vehicle load distribution, velocity, braking rate, battery state of charge, and total braking torque. This hybrid methodology combines the high-accuracy, non-linear mapping of DNNs with the interpretability of DTs. The model was validated in a 14-DOF Simulink environment against two reference strategies (fixed-ratio and baseline) across four different scenarios (UDDS, NYCC, WLTP), including interpolation and extrapolation tests. Key experimental results show the hybrid model accurately tracks the DP-optimal torques (average $��R^2�\approx �0.97�$) and consistently outperforms the reference methods, achieving a 1.26% to 5.06% reduction in net SOC loss. This energy saving translates to a practical gain of 90–383 meters per cycle. Crucially, the model's average inference time of 2.3 ms confirms its computational efficiency and feasibility for real-time implementation on a standard vehicle control unit (VCU).

Accurate traffic flow prediction is essential for the effective management of Intelligent Transportation Systems (ITS). However, traditional methods based on static graph structures often fail to address the complex and nonlinear spatiotemporal dependencies in evolving traffic conditions. To address this challenge, we propose a Dynamic Graph Convolutional Recurrent Network with Temporal Self-Attention (DGCRN-TSA), which integrates a temporal attention mechanism to jointly capture dynamic spatial topologies and long-range temporal patterns. The model incorporates a graph generation module that adaptively learns time-varying adjacency matrices from traffic signals and introduces a trend-aware attention module enhanced by residual-guided decomposition for distinguishing between normal and anomalous traffic behaviours. Experiments on real traffic datasets confirm that DGCRN-TSA achieves superior performance in both short- and medium-to-long-term forecasts. Notably, it reduces MAE by 19.4% on PeMS04 and improves MAPE by 12.2% on PeMS08. The model also ensures high prediction accuracy with strong computational efficiency and an inference speed comparable to AGCRN. DGCRN-TSA offers an efficient and reliable solution for dynamic spatiotemporal modelling and large-scale real-time traffic prediction.

The effective management of connected and autonomous electric vehicle (CAEV) fleets is critical for realising their potential to create safer and more efficient urban transportation systems. As CAEVs increasingly share the road with conventional vehicles, a primary challenge is to optimise their operations within this mixed-traffic reality. This study tackles the joint problem of route selection and charging scheduling for multiple CAEVs by developing a multi-objective optimisation model. The framework is designed to minimise a holistic cost function comprising energy consumption, travel time, charging service fees and penalties for violating user-specified time windows. Critically, the model incorporates real-world complexities, including time-of-use electricity pricing, partial charging strategy and the stochastic queuing delays introduced by non-connected vehicles at charging stations. We adapt the non-dominated sorting genetic algorithm III (NSGA-III) to efficiently solve this complex optimisation problem. The proposed model is demonstrated through a case study using an actual road network from Beijing, China. The results indicate that the proposed model can reduce charging service fees by up to 96% and travel time by 33% when compared to conventional scheduling methods. These findings can offer significant insights for policymakers and platform operators aiming to formulate effective CAEV travel and charging policies in heterogeneous traffic environments.