IET Intelligent Transport Systems最新文献

Wireless sensor networks (WSNs) operating in challenging, resource-constrained dynamic environments often struggle to address the persistent issues related to energy efficiency, node mobility, network coverage and performance. To overcome these research challenges, an innovative hybrid optimisation framework is proposed. This proposed framework effectively integrates squirrel search optimisation (SSO) with adaptive Lévy flights for enhancing the balance between the exploration-exploitation process, and enhanced double Q-learning for adaptive energy-aware routing. In addition, a long short-term memory (LSTM)-based mobility-aware node prediction model enables proactive cluster adaptation and a residual energy-based cluster head (CH) selection process to improve reliability, convergence speed and energy usage. To ensure a uniform workload among sensor nodes, the proposed algorithm incorporates adaptive data aggregation and task-aware load distribution, which minimises the possibility of redundant transmissions and enhances the operational lifespan of nodes under varying node densities. Simulation results across diverse scenarios confirm the effectiveness of our hybrid scheme in terms of performance improvements achieved across various performance evaluation metrics, including a 14.4% improvement in residual energy, a 11.7% improvement in coverage retention, a 18.71% improvement in cluster stability, a 13.42% enhancement in load balancing efficiency, a 4.5% improvement in scalability, a 11.51% enhancement in QoS reliability and a 61% reduction in complexity overhead across various node counts. Additionally, the proposed framework maintains superior network stability and outstanding packet delivery reliability under varying node densities when validated with state-of-the-art algorithms. These capabilities make our hybrid framework a reliable solution for diverse WSN applications where adaptability and resource efficiency are critical priorities.

The multi-agent reinforcement learning (MARL) has been used to control traffic lights to mitigate the traffic congestion problem of urban cities. However, an agent in such algorithm can only have local information of the intersection to which it belongs instead of global information of all intersections that typically cannot be effectively and completely shared by all the agents. Hence, an effective algorithm, which aims to share information between agents at different neighbor intersections to further enhance the performance of MARL in solving the traffic light control problem, will be presented in this study. The proposed algorithm is a two-step communication mechanism that enables agents to share current local information with each other, thereby further improving the performance of MARL for traffic light control plans. To evaluate the performance of the proposed algorithm, we compare it with other state-of-the-art message-passing-based algorithms for solving the traffic light control optimization problem on the simulation of urban mobility (SUMO) simulator. The results show that the proposed algorithm is able to provide better results than state-of-the-art message-passing-based algorithms for the grid, Monaco, and Kaohsiung maps.

In order to solve the problem of excessive model parameters and low real-time performance in driver distraction driving detection tasks, this work proposes a detection model based on cascaded convolutional network and attention mechanism. The model adopts a two-stage architecture. In the first stage, the pre-trained MobileNet is used as the backbone network for basic feature extraction to achieve efficient image feature extraction and significantly reduce the computational complexity. In the second stage, the basic features extracted in the first stage are enhanced by combining the Cascaded ResNet structure with the spatial attention mechanism, so as to improve the capture ability of key features. Finally, the features extracted in the two stages are fused to complete the driver's distraction behavior recognition. The experimental results on the public datasets American University in Cairo (AUC) Distracted Driver and StateFarm Distracted Driver (SFD) show that the proposed model achieves the recognition accuracy of 95.72% and 99.87%, respectively, which is significantly better than the existing mainstream methods while maintaining a low number of parameters. The model has low parameter quantity, high detection accuracy and high real-time performance.

Train platoon improves railway efficiency by coordinating train speeds and inter-train distances. However, actuator delays pose a major challenge to maintaining safe dynamic spacing. This study develops a safety-oriented framework that integrates hybrid $��H_2�/�H_{\infty }�$ control with reachable set estimation to explicitly compute the minimum safe separation under actuator delays. A unified modelling strategy is proposed that incorporates actuator delays together with real-time acceleration disturbances of the leading train. By using forward reachable set analysis, the effect of actuator delays on position errors is estimated and incorporated into the controller to compensate for such delays, thereby improving the safety and robustness of train platoon tracking. The influence of actuator delays on safety during emergency braking scenarios is evaluated through simulation and field experiments. The results show that the forward reachable set of position errors expands as the actuator delays increase. The adoption of $��H_2�/�H_{\infty }�$ controller can reduce the influence of actuator delay on the safety margin by approximately 60%. Compared with the method of eliminating delay using the Lyapunov–Krasovskii functional method, the proposed method ensures the safety of the tracking distance of the train platoon.

Pedestrian crossing prediction, which involves anticipating whether a pedestrian will cross the street or not, is a crucial function in autonomous driving systems. This is also a safety requirement for the interaction of highly automated vehicles and pedestrians. The endeavours in this research domain heavily rely on processing videos captured by the frontal cameras of autonomous vehicles using advanced computer vision techniques and deep learning methods. While recent studies focus on the model architecture for crossing prediction by utilising pre-trained visual feature extractors, they often encounter challenges stemming from inaccurate input features such as pedestrian body pose and/or scene semantic information. In this study, we aim to enhance pose estimation and semantic segmentation algorithms by using synthetic data augmentation (SDA) and domain randomisation (DR) techniques. SDA allows for automatic annotations through predefined agents and objects in a simulated urban environment. However, it creates a domain gap between synthetic and real-world data. To tackle this, we introduce a DR technique to generate synthetic data mimicking various weather and ambient illumination conditions. We evaluated two training strategies on six algorithms for both pose estimation and semantic segmentation algorithms, and ultimately, we target four deep learning architectures for crossing prediction, including convolutional, recurrent, graph, and transformer neural networks. The proposed technique improves the extraction of pedestrian body pose and categorical semantic information, which in turn enhances the state-of-the-art. This results in effective feature selection as the input for the PIP task, improving prediction accuracy by 3.2%, 4.2%, and 6.3% to reach 87.6%, 92.2%, and 73.6% against the JAAD, PIE, and FU-PIP datasets, respectively. The study indicates that using a simulated environment with structural randomised properties can enhance the resilience of the pedestrian crossing prediction to variations in the input data.

The massive real-time data shared by Internet users provides a potentially rich resource for detecting traffic congestion. Targeting China's predominant social network platform, ‘Sina Weibo’, this paper proposes a hybrid deep learning method to mine valuable freeway traffic congestion information. Specifically, original microblog data is extracted and filtered via a customised web crawler coupled with geographical anchors. Afterwards, the selected microblogs undergo rigorous preprocessing, wherein a domain-specific Word2Vec model is trained to represent textual information as high-dimensional word embeddings. To effectively identify congestion-related microblogs, this study develops a ConvBILSTM model that integrates a TextCNN layer for capturing local textual features and a BILSTM layer for modelling global context dependencies. Extensive experimental evaluations demonstrate the superiority of the proposed method compared to benchmark approaches, achieving a recall of 0.8519 and an F1-score of 0.8415. Furthermore, the congestion-prone locations extracted from congestion-related microblogs based on Document Frequency scores are highly consistent with ground-truth data. Overall, this research facilitates timely and accurate reporting of traffic congestion, providing a valuable supplement or alternative to conventional freeway traffic surveillance methods.

Digital twin (DT) plays a vital role across various applications, notably in electric vehicles (EVs). It serves as a virtual counterpart to physical systems. Repurposing legacy EV propulsion systems—those developed prior to the rise of connected vehicle technologies—can reduce electronic waste and support sustainability goals. However, adapting DTs in such systems is challenging due to limited connectivity, incomplete schematics, and performance degradation over time. This paper presents a hybrid-driven DT modelling framework for an EV repurposed with a legacy propulsion system, where some components, like the drive system, have uncertain parameters. A physics-based model is developed for the motor, leveraging its well-defined parameters, while a data-driven model is applied to the drive system due to its uncertainty. The data-driven model was developed using a nonlinear autoregressive neural network with exogenous inputs (NARX). It was trained on physical test bench data and achieved a validation RMSE of 0.04 on unseen data. A hybrid-driven model, combining the NARX-based drive system with a physics-based motor model, was first validated offline in MATLAB/Simulink, then deployed on a speedgoat baseline target machine for real-world testing. The deployment enabled validation under real world vehicle conditions beyond the test bench and assessment of its real-time capability. Real-time testing demonstrated high steady-state accuracy and reliable performance, with an average execution cycle of 8 ms, 60% CPU load, and 300 MB memory usage. Communication via user datagram protocol confirmed the model's real-time readiness and suitability for practical DT integration.

The aim of the study was to analyse whether driving behaviour differs by gender. It focused on two groups of drivers: those under the age of 24 (young drivers) and those who have held their driving licences for less than five years (inexperienced). By examining behaviours such as speeding, driving under the influence of alcohol, seatbelt usage and the use of child-resistant systems, the study sought to gain insights into the prevalence and patterns of these behaviours. To achieve the study's objective, a survey was utilised to gather self-reported behaviour data from 220 young drivers and 271 inexperienced drivers. The frequencies of selected behaviours were analysed, and gender disparities were identified using the Mann-Whitney test and logistic regression analysis. The study demonstrates that gender is a statistically significant factor influencing behaviour primarily among inexperienced drivers and reveals gender-specific driving behaviours among young and inexperienced drivers. Priority actions should focus on reducing speed limit violations among inexperienced males on main roads, restraining alcohol consumption while driving among inexperienced males on urban roads, decreasing phone use for texting and social networking among females on urban roads, and promoting the use of child-resistant systems on both urban and regional roads.

It is a critical problem to improve safety for vehicle platooning systems. This article mainly discusses two issues that affect the safety of the controller. One issue is the communication safety of the controller. When the inter-vehicle distance is larger or smaller than the desired inter-vehicle distance, the system has different requirements for safety and interference reduction. It is well known that the reduction of triggered times can reduce the interference of the vehicle platooning system, further driving safety and communication frequency reduction are contradictory. For the interference reduction, an event-triggered model predictive control (MPC) method with asymmetric design is presented to dramatically reduce the triggered times when the vehicle is in the safe area, and slightly or even not reduce the triggered times when it is in the dangerous area. The other issue is physical safety; if the controller is improperly designed, the vehicle is at risk of rear-end collisions. Thus, the asymmetric weighting error MPC method is presented to design the controller more concerned with the input energy optimization when the vehicle is in the safe area, and pay more attention to the safety when it is in the dangerous area. Further, the accelarating/braking penalty MPC method is presented to avoid the frequent braking when the vehicle goes uphill and the frequent speeding up when it goes downhill. Both methods keep an enough minimal inter-vehicle distance in the transient process of control to avoid rear-end collision. Both issues are solved by the asymmetric designed method, and simulation results are provided to verify the effectiveness and advantages of the proposed methods in both information and physical safety.

With the rapid development of the global aviation industry, multi-airport systems have emerged as a critical component of large urban clusters and regional aviation networks. However, the complexity and uncertainty of air traffic flows in such systems are significantly increased by factors such as weather conditions, emergencies and the intricate interplay of arrival and departure routes across multiple airports, compounded by the complex structure of airspace. To address the challenges posed by the complex and dynamic air traffic flows within multi-airport systems, in this paper, we have introduced a trajectory recognition method based on a new 3D multi-feature trajectory compression (3D-MFTC) representation and clustering. First, a grid sparsity-based approach is proposed to detect and remove abnormal trajectories in multi-airport systems. Then, a novel 3D-MFTC is developed, which employs normalised Euclidean distance to compress 3D trajectory data and adjusts trajectory feature points based on a normal distribution. Then the fast-DTW algorithm is applied to calculate the trajectory similarity of the compressed data. Finally, DBSCAN is utilised to cluster the trajectory within the multi-airport system, with the optimal parameter combinations determined through K-distance graph analysis and grid search. Experimental results demonstrate that the proposed method significantly enhances the accuracy of trajectory similarity computation, enables fine-grained identification of trajectory patterns in multi-airport systems and outperforms traditional clustering algorithms in terms of both clustering performance and visualisation quality.