IET Intelligent Transport Systems最新文献

This paper presents a comprehensive, data-driven framework for personalised driving risk assessment, designed to enhance driver safety within intelligent transportation systems. By leveraging crowdsensed telemetric and road environment data, the framework captures diverse driving behaviours and contextual factors to provide real-time, individualised risk insights. The two-phase framework combines Gaussian Mixture Model (GMM) clustering, Deep Embedded Clustering (DEC), and Fully Connected Network (FCN) for accurate risk classification and prediction, while Deep Q-Learning (DQN) delivers adaptive feedback that encourages safer driving practices. Extensive evaluation shows that our approach outperforms traditional models in both accuracy and adaptability with an accuracy score of 95% and an average F1-score of 0.94, demonstrating its value in capturing complex driver behaviour patterns and contributing a scalable solution for transportation safety.

To improve transportation efficiency, an adaptive speed control method is proposed for ship formation control when a ship formation enters a port with tidal elevation variations. The nonlinear model predictive control (NMPC) method and leader‒follower structure are utilised for the formation keeping and trajectory tracking tasks. The proposed method establishes a ship manoeuvring model and a dynamic speed constraint model for adaptive speed control. A safe distance model is constructed to maintain a safe distance between ship formation members. The proposed safe distance model utilises a Serret‒Frenet (S‒F) coordinate system to describe the positions of ship formation members. Simulation experiments are applied to the North Channel of the Yangtze River. The experimental results indicate that the maximum actual draught accounts for 101.4% of the maximum safe draught without speed constraints. The draft ratio decreases to 99.2% after the adaptive speed control method is applied. This method can be utilised to effectively control ship formation navigation considering variations in tidal elevation.

The continuous expansion of the high-speed rail (HSR) network not only shortens tourists' travel time but also significantly impacts the network attention of destinations. This study uses door-to-door HSR travel times from the Baidu Map API to compute weighted average travel time (WATT) for transportation accessibility (TA) and Baidu Index search data for tourism network attention (TNA) and applies coupling coordination degree (CCD) and relative development degree (RDD) models to evaluate TA-TNA coordination across 28 scenic areas and their host cities in the urban agglomerations in the middle reaches of the Yangtze River (UAMRYR) for 2016–2023. The results indicate that WATT fell by 14.9%, whereas TNA rose overall but remained uneven. The CCD-RDD analysis reveals that most scenic areas exhibit a TA lag category, whereas cities perform better than scenic areas in the coordinated development. To translate these findings into practice, three priorities emerge. (1) Last-mile transport and visitor services in fringe nodes should be improved; (2) Digital marketing and pricing should guide scenic area operations; (3) National and regional transport-tourism governance tools need to be strengthened. These insights provide a quantitative basis for aligning rail expansion, destination marketing, and infrastructure finance to achieve balanced regional tourism growth.

The transition to electric buses (e-buses) is essential for reducing greenhouse gas emissions in urban transit systems. However, successful e-bus deployment requires careful planning to ensure service reliability while minimising costs. A key challenge in this transition is determining the replacement factor, the ratio of e-buses needed to replace the current diesel-engine bus fleet for a certain route. This factor is essential for transit agencies as it directly influences fleet size, capital investment, and operational efficiency. Accurately estimating replacement factors allows agencies, to prioritise routes where electrification achieves the highest economic and environmental benefits while preventing unnecessary fleet expansion and idle capacity by selecting routes with low replacement factors. This study develops a framework for estimating e-bus replacement factors based on route characteristics, vehicle attributes, and external conditions. Multiple machine learning models are evaluated, with XGBoost achieving the highest accuracy (R² = 0.93). Model interpretability using SHapley Additive exPlanations (SHAP) analysis identifies the average bus speed and ambient temperature as the main variables affecting the replacement factor. The proposed framework enables transit agencies to optimise fleet deployment by prioritising routes with lower replacement factors, maximising e-bus utilisation, and achieving cost efficiencies while aligning with environmental objectives.

Electric vehicles (EVs) provide significant advantages for sustainable transportation, such as reduced energy consumption, the ability to integrate with renewable energy sources, and emission reductions. Nevertheless, range anxiety, high battery costs, and long charging times limit the adoption of EVs. Accurately estimating driving range is one of the solutions to overcome these limitations. This study proposes a method that combines an extra tree regressor (ETR) model and an artificial rabbit optimization (ARO) algorithm to predict the driving distance using a comprehensive dataset for EVs. In our experiments, we compared ARO with well-known hyperparameter optimization methods such as grid search (GS) and random search (RS), and tested the models across multiple train and test splits. Besides using the complete feature set, we applied recursive feature elimination (RFE) to select an informative subset and re-evaluated all methods. With all features, the best configuration of the proposed algorithm achieved an R-squared (R²) of 0.84, a root mean square error (RMSE) of 14.38, a mean absolute error (MAE) of 7.70, and a mean squared error (MSE) of 220.12. Using the selected subset of seven features, the proposed model reached an R² of 0.84, with an RMSE of 14.88, an MAE of 6.75, and an MSE of 221.53. Finally, the contribution of each feature's to the predicted driving range was analysed using shapely additive explanations (SHAP). The findings of the study emphasize the value of integrating machine learning (ML) models and hyperparameter search methods into electric vehicle range-estimation systems to improve driver confidence and support sustainable transportation.This study advances the current understanding of range prediction and contributes to reducing range anxiety, thereby supporting extensive adoption of EVs. The findings of the study indicate that the integration of ML approaches in the range estimation of EVs can play a critical role in increasing driver confidence and supporting sustainable transportation. This study contributes to the existing knowledge in the field of range estimation and is an important step toward the broader adoption of EVs.

Large-volume and heavy-duty transports are essential for the successful and timely execution of large-scale industrial, socio-political and climate-related projects. As these transports grow in size and complexity, the planning process becomes increasingly challenging for all stakeholders involved. To overcome these challenges, detailed planning processes are required, especially for the trafficability of narrow passages along the transportation route. This paper introduces an advanced methodology for an enhanced 3D trafficability analysis with collision detection of large-volume and heavy-duty transports. By employing high-resolution, dense, colored 3D point clouds alongside detailed transport models, this approach offers a more accurate and comprehensive assessment of the feasibility of the transport. The methodology is further generalized to accommodate a wide variety of transport configurations and maneuvers, allowing for automated analysis across different scenarios. The primary contribution of this research lies in its ability to significantly improve collision detection accuracy and provide detailed visualizations, thereby optimizing the digital planning process of large-volume and heavy-duty transports. The findings demonstrate a distinct advantage of the 3D trafficability analysis over traditional 2D methods, especially in complex environments, leading to cost-effective and reliable transportation planning.

Negative externalities refer to costs arising from transportation activities that are not borne by service providers or consumers, often leading to their neglect in freight transportation planning. This study proposes a novel framework for optimising water-road transportation route selection by incorporating two key negative externalities: fuel consumption and traffic risk. Traffic risk is assessed using a safety performance function, while fuel consumption is estimated based on the Handbook of Emission Factors for Road Transport. The proposed framework is applied to California's road network and port system, where the optimal operation area for each major port is determined and compared across different optimisation objectives: trip distance, fuel consumption and traffic risk. Results indicate that the optimal operation area varies significantly depending on the relative weight assigned to each objective. The findings demonstrate that optimising routes beyond just minimising distance can reduce fuel consumption and traffic risk, highlighting the substantial differences in optimal operational areas under different criteria.

With the rapid development of model cities, urban rail transit (URT) systems have emerged as a crucial component in urban public transport, and passenger flow prediction serves as the cornerstone for planning the travel, avoiding the congestion, and improving the travel efficiency. In order to predict short-term passenger flow of the URT system, a hybrid convolutional neural network (CNN)-long short-term memory (LSTM)-particle swarm optimisation (PSO) model is proposed to accommodate both the spatial and temporal features of passenger flow. First, spectral clustering is employed to extract four different types of stations, in which the Calinski–Harabasz (CH) index is considered. Second, the hybrid CNN-LSTM-PSO model is constructed to predict the short-term passenger flow for different types of stations, in which CNN can extract the abstract feature with a multi-layer convolutional structure, LSTM can deal with time series data, and the PSO algorithm is employed to optimise some parameters. Third, the data from Hangzhou urban rail transit in 2019 are employed for prediction. The results show that the proposed hybrid model reveals the best performance in accuracy by comparing the equivalent CNN-LSTM, LSTM and autoregressive integrated moving average (ARIMA) models. At last, some empirical suggestions are provided to benefit both the passengers and operation and management departments of the URT system.

The optimization of spatial and temporal resources at signalized intersections is a critical aspect of intelligent transportation systems. Traditional traffic signal control methods, which usually rely on fixed signal timings and lane assignments, are suboptimal in addressing changing traffic conditions. Additionally, understanding traffic flow in a large scale is often challenging due to the lack of traffic flow monitoring infrastructure. This paper introduces Trans-Space, a novel framework that leverages transfer learning and space computing for managing spatiotemporal traffic resources at signalized intersections. Trans-Space consists of two core modules: (space computing for optimized traffic system (SCOTS) and traffic optimization with spatial–temporal control agents (TOSCA). SCOTS configures satellite constellations for high-resolution Earth observation images and utilizes space computing to extract real-time traffic flow parameters. TOSCA employs hierarchical reinforcement learning agents to optimize lane directions and signal timings based on the data provided by SCOTS. TOSCA incorporates knowledge transfer that adapts traffic management strategies from source to target intersections. Through extensive simulations, Trans-Space demonstrated superior performance over traditional and state-of-the-art models in traffic flow metrics. The paper concludes with a discussion on the prospects of space computing in traffic management and potential directions for future research.

The study investigates emergency evacuation strategies for high-capacity multi-line transfer metro stations, with a focused examination on congestion dynamics induced by interweaving passenger flows. Taking a three-line interchange station in Xi'an, China, as an example, a 3D emergency evacuation physical model was established using AnyLogic software, with pedestrian parameters and behavioural logic for security checks, level transfers, and evacuation configured through Java programming. Observations from a scenario involving 2200 passengers revealed that safety exits B, C and E, along with escalator groups 1 and 4 in high-traffic areas, were the station's evacuation bottlenecks, leading to congestion and stampede risks. Pedestrians tended to choose the nearest exits, resulting in a peak density of up to 3.79 persons/m². To address these challenges, this study proposed two optimised evacuation routes. After optimisation, evacuation time was significantly reduced by over 10%, meeting safety requirements. These findings contribute to improving emergency evacuation strategies for complex multi-line subway stations.