Railway turnout system is a key infrastructure to railway safety and efficiency. However, it is prone to failure in the field. Therefore, many railway departments have adopted a monitoring system to monitor the operation status of turnouts. With monitoring data collected, many researchers have proposed different fault-diagnosis methods. However, many of the existing methods cannot realize real-time updating or deal with new fault types. This paper—based on imbalanced data—proposes a Bayes-based online turnout fault-diagnosis method, which realizes incremental learning and scalable fault recognition. First, the basic conceptions of the turnout system are introduced. Next, the feature extraction and processing of the imbalanced monitoring data are introduced. Then, an online diagnosis method based on Bayesian incremental learning and scalable fault recognition is proposed, followed by the experiment with filed data from Guangzhou Railway. The results show that the scalable fault-recognition method can reach an accuracy of 99.11%, and the training time of the Bayesian incremental learning model reduces 29.97% without decreasing the accuracy, which demonstrates the high accuracy, adaptability and efficiency of the proposed model, of great significance for labor-saving, timely maintenance and further, safety and efficiency of railway transportation.

Demand-responsive transportation (DRT) is a flexible passenger service designed to enhance road efficiency, reduce peak-hour traffic, and boost passenger satisfaction. However, existing optimization methods for initial passenger requests fall short in addressing real-time passenger needs. Consequently, there is a need to develop real-time DRT route optimization methods that integrate both initial and real-time requests. This paper presents a two-stage, multi-objective optimization model for DRT vehicle scheduling. The first stage involves an initial scheduling model aimed at minimizing vehicle configuration, and operational, and CO₂ emission costs while ensuring passenger satisfaction. The second stage develops a real-time scheduling model to minimize additional operational costs, penalties for time window violations, and costs due to rejected passengers, thereby addressing real-time demands. Additionally, an enhanced genetic algorithm based on Non-dominated Sorting Genetic Algorithm-II (NSGA-II) is designed, incorporating multiple crossover points to accelerate convergence and improve solution efficiency. The proposed scheduling model is validated using a real network in Shanghai. Results indicate that real-time scheduling can serve more passengers, and improve vehicle utilization and occupancy rates, with only a minor increase in total operational costs. Compared to the traditional NSGA-II algorithm, the improved version enhances convergence speed by 31.7% and solution speed by 4.8%. The proposed model and algorithm offer both theoretical and practical guidance for real-world DRT scheduling.

In the face of rapid urbanization and the increasing number of vehicles, urban centers are struggling with traffic congestion. This study presents a dynamic travel strategy using the MATSim platform to schedule urban travel, incorporating a model for shared autonomous electric vehicles. The model is evaluated using a baseline scenario for Shanghai, exploring the effects of vehicle range, charging capabilities, and power supply strategies on the uptake of shared autonomous electric vehicles. Results indicate that enhancements in vehicle range and charging efficiency slightly decrease the use of autonomous vehicles by 2.5%, as the existing vehicle specifications already meet daily travel needs in Shanghai. Additionally, the transition from traditional charging stations to a battery-swapping system does not significantly alter overall travel behavior of shared autonomous electric vehicles. These findings provide insights into the deployment of intelligent traffic systems to alleviate urban traffic congestion.

In order to improve the frequent Doppler bias in the high-speed mobile scenario, improve the communication quality on HSR (High-Speed Railway). An algorithm can improve communication performance through RIS (intelligent metasurface) phase optimization. The optimization process is modeled as a reinforcement learning task, where the agent in the channel learns to select the optimal RIS phase by interacting with the environment to reduce the effect of Doppler shift.The model covers characteristic parameters such as signal propagation path and frequency related to RIS, and the agent influences the beam direction by adjusting the phase, and learns the optimal phase configuration in the high-speed moving scene.The RIS phase optimization method implemented based on the deep reinforcement learning algorithm has high convergence and has significant advantages in improving the Doppler frequency shift problem. Compared with the traditional frequency bias estimation method, this method can significantly improve the transmission performance. The effectiveness of the proposed algorithm is verified by performing large-scale training and evaluation in the simulation environment.

[Objective] To address blurred object boundaries, difficult small-target localization, and unstable bounding-box regression caused by low-quality samples in infrared images of substation electrical equipment, an infrared object detection method based on YOLOv8n improved with Wise-IoU is proposed. [Method] The lightweight YOLOv8n is adopted as the baseline detector, and the original bounding-box regression loss is replaced by Wise-IoU without changing the inference structure. By introducing the outlier degree and a dynamic non-monotonic focusing mechanism, Wise-IoU adaptively adjusts the gradient contribution of samples with different qualities, emphasizes ordinary-quality samples with effective learning value, and suppresses redundant optimization of high-quality samples as well as harmful gradients from low-quality samples. [Result & Conclusion] Experiments on a self-built infrared image dataset of 12 classes of substation electrical equipment show that the YOLOv8n-WIoU model achieves 81.6% mAP@0.5 on the test set, outperforming the baseline YOLOv8n by 6.5 percentage points. Since Wise-IoU only affects the training loss and does not change the network structure during inference, the proposed method improves localization robustness while maintaining the lightweight and real-time characteristics of YOLOv8n.