In existing urban rail systems, most trains are operated by automatic control, which places high demands on the control effectiveness of automatic train operation (ATO). In this study, a train operation model considering the response delay is firstly constructed. Subsequently, by analyzing and comparing the existing mainstream research methods, neural network and sliding mode control techniques were selected and incorporated into the speed control of the train. Among them, the dynamic sliding mode technique, is used to optimize the PID control effect of ATO. Single neuron and back propagation (BP) neural network algorithms are applied to the selection of PID control parameters. The study selects the difference between the actual speed of the control train and the target speed as the control objective. Through continuous optimization and iteration of the control parameters, the control accuracy of the train operation was improved. Finally, this study validates and simulates different model control methods using the actual operation data of urban rail transit. The results show that the sliding mode PID control model optimized with BP neural network performs better in terms of error distribution, average error value, and control effect variance under different simulation scenarios, showing good tracking ability and robustness. The related research results are expected to be applied to the initial selection of rail vehicle control parameters, dynamic adaptive operation control, and other fields, providing practical help to rail operators and rail signaling companies.

This paper proposes a new hybrid hierarchical roadside unit (RSU) deployment algorithm that combines traditional RSU deployment strategies with temporary RSUs (t-RSUs) selected from many parked vehicles. The driving vehicles in the corresponding RSU coverage section are divided into clusters and the demand is thus summarized. We first solve the problem of how to choose the appropriate cars serving as t-RSUs and then optimize the RSU deployment with the dynamic existing t-RSUs. We simulate the deployment application in a downtown area of Shanghai, China, using the simulation of urban mobility (SUMO), version 1.11.0, an open-source traffic simulation package, and its traffic-control interface (TraCI) with Python. The simulation results show that, within a given economic cost, the proposed hybrid hierarchical RSU deployment algorithm outperforms the maximum vehicle coverage deployment algorithm (MVCD) and the random deployment algorithm (RandDeploy) in terms of overall vehicle coverage and point-to-point vehicle connectivity ratio. Our scheme provides new solutions and ideas for the deployment of RSUs in future urban traffic environments.

In order to enhance the precision of network traffic prediction for multi-area vehicle networks, this paper proposes a two-tier distributed Software-Defined Vehicular Network (SDVN) architecture equipped with multiple controllers, which is subsequently deployed to manage traffic across regions, thus minimizing communication costs and enabling seamless vehicle movement. We firstly build on control plane placement (CPP) research, focusing on deployment strategies that impact scalability, performance, and fault tolerance. It highlights the importance of hierarchical decision-making with multiple controllers handling varying traffic demands. A comprehensive comparison of various machine learning and deep learning algorithms is then conducted to evaluate their efficacy in forecasting SDVN traffic patterns, which is crucial for system efficiency. Experiments show the proposed architecture’s effectiveness in traffic prediction and management in Shenzhen’s Longhua New District. The study confirms that the SDVN system enhances traffic prediction and management, improving urban mobility and Internet of Vehicles. The paper offers a framework using advanced technologies to address challenges in traffic prediction and management in modern vehicular networks.

The rapid proliferation of heterogeneous applications, from latency-critical video delivery to bandwidth-intensive file transfers, poses increasing challenges for modern communication networks. Traditional traffic engineering approaches often fall short in meeting diverse Quality of Experience (QoE) requirements under such conditions. To overcome these limitations, this study proposes a QoE-driven distribution framework for mixed traffic in Software-Defined Networking (SDN) environments. The framework integrates flow categorization, adaptive path selection, and feedback-based optimization to dynamically allocate resources in alignment with application-level QoE metrics. By prioritizing delay-sensitive flows while ensuring efficient handling of high-volume traffic, the approach achieves balanced performance across heterogeneous service demands. In our 15-RSU Mininet tests under service number = 1 and offered demand = 10 ms, JOGAF attains max end-to-end delays of 415.74 ms, close to the 399.64 ms achieved by DOGA, while reducing the number of active hosts from 5 to 3 compared with DOGA. By contrast, HNOGA exhibits delayed growth of up to 7716.16 ms with 2 working hosts, indicating poorer suitability for latency-sensitive flows. These results indicate that JOGAF achieves near-DOGA latency with substantially lower host activation, offering a practical energy-aware alternative for mixed traffic SDN deployments. Beyond generic communication scenarios, the framework also shows strong potential in Intelligent Transportation Systems (ITS), where SDN-enabled vehicular networks require adaptive, user-centric service quality management. This work highlights the necessity of coupling classical traffic engineering concepts with SDN programmability to address the multifaceted challenges of next-generation networking. Moreover, it establishes a foundation for scalable, adaptive data distribution strategies capable of enhancing user experience while maintaining robustness across dynamic traffic environments.

Virtual coupling is a key technology for enhancing rail transit efficiency, enabling trains to coordinate closely via wireless communication. However, conventional control strategies, particularly model-based ones, face significant challenges such as model inaccuracies, insufficient robustness, and high computational overhead. These limitations are particularly pronounced in scenarios with large, abrupt changes in control objectives, most notably during the coupling and decoupling transitions of train convoys. To address these limitations, this paper proposes an adaptive, model-free control framework based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. By learning optimal control policies directly from environmental interaction, this Deep Reinforcement Learning (DRL) approach bypasses the requirement for a precise system model. Comprehensive evaluations across representative scenarios demonstrate a nuanced, task-dependent performance profile, where the method’s effectiveness varies with the operational context. While the TD3 controller exhibited a trade-off between speed and transient precision during coupling, it significantly outperformed baseline methods in the more complex decoupling scenarios, demonstrating superiority in both operational efficiency and tracking accuracy. Across all scenarios, the proposed method demonstrated superior adaptability and robustness, reducing the average completion time by 14.9% compared to Sliding Mode Control (SMC) and 16.2% compared to Model Predictive Control (MPC), while also decreasing the average tracking error by 13.4% and 24.5%, respectively. In conclusion, the proposed TD3-based framework presents a viable and high-performing solution for the complex dynamics of virtual coupling transitions. It demonstrates the significant potential of model-free DRL to enhance the operational efficiency, precision, and adaptability of future intelligent rail transit systems.

Virtual coupling (VC) technology has been introduced in railway systems to enhance operational capacity. One of the key challenges in VC is ensuring a safe and effective train control strategy. For VC convoy operation control, while the existing literature mainly utilizes a distributed control architecture that focuses on the performance of individual trains, limited research focuses on optimizing the control trajectories for the entire convoy. To address this gap, this paper aims to plan a cooperative train trajectory for a VC convoy to improve the overall performance with respect to safety, punctuality, energy efficiency, and so forth. A hierarchical hybrid architecture is proposed for strategy planning and control of a VC convoy, considering the overall benefits of the convoy based on distributed control. Different initial conditions and the transition process of train states at junctions are investigated. A multi-objective nonlinear integer programming model is developed to optimize cooperative train trajectories during the coupling process by constructing a discrete speed–time grid network via the state lattices method. Simulation results demonstrate that the proposed method consistently meets the real-time safe separation requirements, showing adaptability to various combinations of initial train conditions and train operation plans. Compared with the two existing methods, the proposed method limits the deviation between the actual and the target operation times within 0.4 s, while reducing traction energy consumption by 2–4%, respectively. The proposed strategy advances the state of the art in VC train convoy control and shows promising potential to be applied to a real-world operational environment.