IET Intelligent Transport Systems最新文献

How to accurately process and achieve good transient performance in a short period of time is a key consideration factor for the system. A hyperbolic sine function is used to construct a novel predefined-time convergent prescribed performance function. This algorithm introduces a set of new predefined standards for time convergence assessment based on gamma functions and Riemann zeta functions. By integrating performance indicators of speed, stability and efficiency into the design of the prescribed performance function, the performance framework ensures the achievement of establishing a comprehensive performance optimization model. The upper limit of the settling time is studied, and sufficient conditions for achieving predetermined time convergence are established, validated through experiments using unmanned surface vessels.

Autonomous vehicles (AVs), driven by state-of-the-art deep learning and computer vision technologies, can revolutionize current mobility systems in modern transportation. Driverless AVs are slowly integrated into public transportation with significant advantages for the passengers and public transport operators. However, passenger safety and comfort are two of the main challenges that need to be addressed. This work presents a complete in-cabin monitoring framework with a suite of services, employing deep learning algorithms using a variety of onboard sensors at the edge. This proposed framework offers various innovative services aimed at enhancing security, monitoring passenger presence, accommodating diverse needs, and personalizing the passengers' travel experience, while also reducing the workload of human safety officers. Experimental results demonstrate the framework's effectiveness in identifying abnormal events with a high accuracy, employing multiple datasets and custom in-cabin scenarios. Additionally, the system effectively conducts automated passenger counting and facial identification, ensuring real-time responsiveness under diverse operational conditions. Overall, the novelty of this work lies in the framework's multimodal approach, integrating visual and audio analysis, to achieve robust performance across various scenarios, significantly contributing to the advancement of autonomous driving technologies.

Maintaining an appropriate distance between trains is key to the normal operation of multiple trains in the virtual coupling mode. However, owing to physical limitations, the saturation of the control system is prone to occur during actual train operations, which makes it difficult to maintain a safe distance between adjacent trains when the speed changes. An anti-saturation sliding mode control algorithm for multiple virtually coupled trains was proposed to address this issue. First, according to the virtual coupling dynamics model of multiple heavy-haul trains (HHTs), an improved finite-time anti-windup compensator (FAWC) suitable for the train model was designed such that the compensation factor rapidly converged within a finite time. Second, the FAWC was introduced into the controller to suppress the input saturation phenomenon of trains. Then, a finite-time dual anti-saturation sliding mode controller (FDA-SMC) was constructed based on the barrier Lyapunov function in combination with the sliding mode algorithm against input constraints to suppress the impact of input and output saturation on the tracking accuracy for the relative position between adjacent HHTs. The stability of the closed-loop system was verified using the Lyapunov stability theory. Finally, the simulation and experimental results showed that the proposed algorithm demonstrated advantages in terms of anti-saturation performance and maintained a safe distance between adjacent HHTs.

As automated driving technology advances, the role of the driver to resume control of the vehicle in conditionally automated vehicles becomes increasingly critical. In the SAE level 3 or partly automated vehicles, the driver needs to be available and ready to intervene when necessary. This makes it essential to evaluate their readiness accurately. This article presents a comprehensive analysis of driver readiness assessment by combining head pose features and eye-tracking data. The study explores the effectiveness of predictive models in evaluating driver readiness, addressing the challenges of dataset limitations and limited ground truth labels. Machine learning techniques, including LSTM architectures, are utilised to model driver readiness based on the spatio-temporal status of the driver's head pose and eye gaze. The experiments in this article revealed that a bidirectional LSTM architecture, combining both feature sets, achieves a mean absolute error of 0.363 on the DMD dataset, demonstrating superior performance in assessing driver readiness. The modular architecture of the proposed model also allows the integration of additional driver-specific features, such as steering wheel activity, enhancing its adaptability and real-world applicability.

The current method to address the reduced accuracy of target detection algorithms in hazy weather scenes is mainly to first use image dehazing algorithms to restore hazy images, and then input the restored images into target detection algorithms to obtain detection results. However, the images restored by the image dehazing model deviate from real clear images, and do not completely recover the features required by the target detection algorithm, thus limiting the improvement of the detection accuracy of the target detection model. This paper proposes a hazy weather target detection algorithm based on large convolution kernels and knowledge transfer (KTnet). First, a large convolution attention dehazing module is embedded into the backbone network of faster R-CNN to form a dehazing backbone network. Considering the high-dimensional features of the deep backbone network, a lightweight fusion attention module is designed. A loss function is also designed and the adapter model is employed to devise training methods for knowledge transfer and fine-tuning. Extensive experimental results on various hazy weather target detection datasets show that KTnet has achieved significant effectiveness.

Understanding travel patterns and predicting travel destinations has gained significant attention in the field of transportation research. This study proposes a methodology that utilizes continuous hidden Markov models (CHMMs) to estimate activity sequences for each travel chain and employs a travel destination prediction model based on a random forest (RF) model. Furthermore, it explores the optimization of the results from HMM using the latent Dirichlet allocation (LDA) model and applies it in predicting travel destinations. In the experiment, the dataset collected from unique travellers in Seoul city, South Korea, is used to validate the proposed model, which includes time stamps of origin and destination, location, travel mode and transfer nodes. Research findings show that during the modelling phase of the continuous hidden Markov model, the Gaussian mixture model categorizes the feature vectors into eight distinct groups. The estimated membership probability indicates involvement in four different activities. It also explains the relationship between derived activities. Finally, given the observed features, the proposed model provides an effective method for estimating the most likely sequence of activities in the travel chain. The results can help conduct further activity-based traffic demand analysis and improve the service quality of the transportation system.

To address the challenges posed by incomplete data in passenger flow prediction and organizational tasks, this paper proposes ProbSparse self-attention conditional generative adversarial imputation net (ProbSA-CGAIN), a novel imputation model framework built on the enhanced generative adversarial network (GAN). The model leverages conditional GANs for controlled data generation using external conditional information. It adopts a denoising autoencoder structure for reconstructing and estimating missing passenger flow data. The integration of an efficient ProbSparse self-attention mechanism captures spatiotemporal evolution features, reducing computational complexity. Additionally, the model incorporates auxiliary conditional information to enhance data imputation accuracy by learning interdependencies among multiple data variables. Further, the model integrates local positional encoding and multi-layer global temporal encoding, offering diverse perspectives on spatiotemporal information. Experimental evaluations with real passenger flow data demonstrate the model's superiority over advanced baseline models across various missing patterns and rates. Notably, it exhibits high stability in data restoration, particularly for datasets with higher missing rates, affirming its effectiveness in predicting and inferring missing passenger flow data based on auxiliary data and multi-view positional information, ensuring reliable imputation. The experiments also assess the model's proficiency in attributing different spatiotemporal features, confirming its commendable training and restoration efficiency.

Traffic management has become a major issue in urban planning due to the increasing number of vehicles on urban roads. In this study, we introduce a novel approach using the Reinforcement Learning (RL) technique to address the vehicle routing problem (VRP). We explored the effectiveness of Double Q-Learning enhanced by Prioritized Experience Replay (DQL-PER) in optimizing vehicle routing to shorten travel times and reduce congestion. Using the Simulation of Urban Mobility (SUMO), this method manipulates traffic flow during peak hours to improve urban mobility. DQL-PER stands out due to its superior performance in managing complex traffic systems characterized by multiple interconnected variables and dynamic conditions inherent in urban traffic networks. Compared to standard Q-learning, DQL-PER reduces overestimation bias and facilitates faster convergence toward optimal solutions. This paper includes a comparison between DQL-PER and other RL methods, namely Q-learning, Double Q-learning (DQL), and deep Q-network (DQN), demonstrating its benefits through simulations and analysis. We also perform a scalability analysis to evaluate the algorithm's performance across network sizes, with node counts $��N�=��39�,�545�,�1672�,�3236�,��\text{and}��9652��$, showing that DQL-PER performs exhaustively over larger networks, demonstrating its scalability potential. DQL-PER offers a scalable solution with the potential to transform urban transportation systems.

Vehicle trajectory prediction enables autonomous vehicles to better reason about fast-changing driving scenarios and thus perform well-informed decision-making tasks. Among different prediction approaches, deep learning-based (DL-based) methodologies stand out because of their capabilities to efficiently summarise historical data, infer nonlinear behavioural patterns from human driving data, and perform long-horizon prediction. This work reviews the DL-based methods that have shown promising results, organising them in terms of usage of the input data, separating the encodings of the target vehicle's historical data, surrounding vehicle's historical data, and road layout data. In particular, this paper explores the relationships between the scope of the prediction components and the input data formats, as well as the connections with other elements in the same prediction framework, including vehicle interaction and road scene mining. This information is crucial to understand complex architectural decisions and to provide guidance for the design of improved solutions. This work also compares the performance of the most successful prediction models, establishing that appropriate encodings of vehicle interactions and road scenes improve trajectory prediction accuracy, with the best performance achieved by attention mechanism and Transformer-based models. Finally, this work discusses future research directions, including considerations for real-time applications.

Urban traffic control is a complex and dynamic multi-agent challenge, characterized by the need for efficient coordination and real-time responsiveness in fluctuating traffic conditions. Traditional methods often fall short in adapting to these dynamic environments. This article introduces “DDPGAT”, a novel framework that merges Multi-Agent Deep Deterministic Policy Gradients (MADDPG) with Graph Attention Networks (GATs) for optimized urban traffic control, further enhanced by a unique moral reward component. DDPGAT empowers traffic signal controllers as independent agents using GATs for dynamic road importance assessment. Shared attention scores during training enhance each agent's understanding of local and wider traffic patterns, essential for developing adaptive control policies. A key innovation in DDPGAT is the moral reward function, encouraging decisions that consider neighboring intersections' traffic, thus promoting ethical traffic management. The experiments demonstrate that DDPGAT significantly boosts traffic throughput and reduces congestion, confirming its effectiveness in diverse traffic conditions. The integration of MADDPG, GATs, and a moral reward strategy in DDPGAT presents a sophisticated, robust approach for managing the complexities of urban traffic control, marking a notable progression in intelligent traffic system technologies.