In the context of global decarbonisation initiatives, the logistics sector faces dual challenges: its substantial energy consumption and carbon footprint conflict with societal goals for a low-carbon economy, while escalating pressures from urban traffic congestion also inflate distribution costs. The environmental externalities and economic losses induced by the combination of inefficient routing and congestion have jointly motivated the emerging research field of low-carbon vehicle routing optimisation. To reconcile these issues, this study develops a bi-level programming framework for low-carbon-oriented vehicle routing optimisation that explicitly accounts for road congestion. The upper-level model aims to minimise the total cost (including vehicle fixed cost, transportation cost, carbon emission cost and time-window penalty) by internalising a carbon tax constraint. The lower-level model employs a user equilibrium (UE) model, focusing on minimising travel time from the perspective of road users. A hybrid solution methodology (GA-Tent & Frank-Wolfe) is proposed, integrating an enhanced genetic algorithm with Tent chaos mapping for global optimisation and a modified Frank-Wolfe algorithm for traffic assignment. Finally, a case study using the Sioux Falls network demonstrates that traffic congestion increases carbon emissions, but a moderate carbon tax increase can effectively reduce vehicle carbon emissions. These insights suggest policymakers should implement progressive carbon pricing mechanisms coupled with dynamic congestion pricing, while logistics operators should prioritise route optimisation systems with real-time traffic adaptation capabilities.
Frequent traffic crashes on urban roads seriously threaten public safety and traffic operations. Accurate risk prediction is vital for improving management efficiency and developing intervention measures. This paper proposes a traffic crash risk prediction model integrating multi-source heterogeneous data. It constructs a dynamic spatio-temporal graph network (DTGN) based on edge-aware graph convolutional networks (EGCN) and introduces a dynamic threshold risk stratification mechanism and local crash density (LCD) indicators to alleviate the issue of “zero inflation” in low-frequency areas. The model combines graph convolutional and spatio-temporal convolutional networks to extract multi-dimensional spatio-temporal features and enhances the ability to identify high-risk areas through a weighted loss function. The city is partitioned into hexagonal grid units, and a dynamic adjacency matrix is constructed to capture spatial associations and evolutionary features. Experimental results indicate that DTGN performs effectively in processing multi-source data and extracting key risk features, achieving an accuracy rate of 87% in high-risk area predictions, thereby providing more practical early warning support and decision-making basis for urban traffic safety management.
Frequent freeway accidents cause significant casualties and economic losses, necessitating robust risk assessment methods. This study proposes a quantitative method for assessing freeway driving risk using continuous observational data from toll transactions. Based on toll data from the Yongguan Freeway in Guangdong Province, China (June–August 2022), 18 risk characteristic indicators for cargo vehicles and 13 for passenger vehicles were developed. Factor analysis reduced these indicators into five common factors, followed by K-means++ clustering to categorise vehicles into risk groups. The entropy weight method calculated risk scores, determining risk levels. The model identified 17.75% of cargo vehicles as high-risk and 14.03% as moderately high-risk, and 7.47% of passenger vehicles as high-risk and 1.08% as moderately high-risk. Validation using rescue events per 10,000 vehicles (RM) from a Guangdong Province accident database, due to limited crash data availability, confirmed consistency with model-assigned risk levels, supporting targeted safety interventions.
With the rapid development of intelligent operation and management of urban rail transit, accurate passenger flow prediction is crucial for management and operation. However, the complex, nonlinear and non-smooth characteristics make detecting passenger flow evolution features challenging. In this regard, a novel decomposition integration model, IC-TLCA, is proposed. The model first decomposes the raw passenger flow data into multiple sublayers with different frequencies using an improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) algorithm to better capture the intrinsic structure of the data. Then, the temporal convolutional network (TCN), long-short-term memory network (LSTM) and channel attention module (CAM) are combined to predict each sublayer separately. Finally, by combining the prediction results of each sublayer, the final passenger flow prediction is obtained. After comparing with all baseline models and conducting ablation experiments, the IC-TLCA model has been validated to be innovative in theory and superior in practical applications, thereby providing an effective solution for passenger flow prediction.
Traffic flow prediction is an important component of intelligent transportation systems. Most existing studies adopt spatio-temporal methods for real-time short-term prediction. However, when dealing with complex and dynamically changing historical data in long-term prediction tasks, several challenges arise, such as high complexity, high computational cost and unstable spatial dependencies. In response to this deficiency, this paper proposes a trend-season dual-branch fusion network with harmonic weighting for long-sequence traffic flow prediction. This model decomposes the original sequence into trend and seasonal components. Specifically, the trend branch applies multi-scale sampling, and the trend components at different scales are fused in a top-down manner to enhance information representation. Finally, a linear layer is used to predict the trend results. The seasonal branch encodes the seasonal components through wavelet transform, and a harmonic weighting mechanism is designed to adaptively fuse the prediction results of both branches. Experiments conducted on multiple public transportation datasets demonstrate that the proposed model significantly outperforms existing mainstream methods in long-term predictions at time steps of 96, 192, 288 and 336, thereby verifying the effectiveness and robustness of time-series modelling approach in long-sequence prediction scenarios.
