Global warming demands urgent action to reduce greenhouse gas emissions, with road transport being a major contributor. Electrification of passenger cars is widely promoted as a key mitigation strategy; however, its real-world impact remains debated. This study develops a calculation framework to estimate annual fleet-level greenhouse gas emissions based on actual consumption data, vehicle type distribution, age and mileage. Using this method, short- and mid-term drawbacks of electrification are quantified alongside long-term benefits through scenario analysis aligned with the European Union 2035 target. Results indicate that electrification begins to deliver net emission reductions after 2032, achieving a 5.3% decrease by 2035 compared to an internal combustion engine fleet. These findings highlight the temporal dynamics of electrification and provide evidence-based insights for policymakers and industry stakeholders.
In the operation of railway transportation enterprises, having prior knowledge of future loading volumes and trends at freight stations is crucial for optimal deployment of empty cars and the development of daily operational plans. Long-term freight volumes at railway stations often exhibit cyclical patterns influenced by seasonal fluctuations, holidays and other factors. Additionally, changes in freight volume are significantly affected by the proportion of freight from various industries. To address these dynamics, this study proposes a hybrid prediction model for long-term loading volumes at railway freight stations. This model predicts by stage segmentation through peak-valley segmentation (PVS), variational mode decomposition (VMD) and an attention mechanism integrated into a temporal convolutional network (TCN). Using historical freight volumes from the Shuohuang railway freight station as a case study, we employed mean absolute percentage error (MAPE), mean absolute error (MAE) and root mean square error (RMSE) as evaluation metrics to assess the combined predictive performance of the PVS, VMD and TCN methods. The experimental results show that the PVS-VMD-A-TCN model significantly improves the accuracy of long-term freight volume predictions. Compared to traditional methods such as ARIMA, GRU and TCN, it exhibits superior predictive performance, offering a new approach for accurately forecasting long-term loading volumes.
With the improvement of deep learning technology, traditional obstacle avoidance approaches for blind people can no longer meet practical needs. In response to the problems of poor adaptability to multiple scenarios and low obstacle avoidance rates in current obstacle avoidance methods for blind people, this study proposes an intelligent obstacle avoidance method for blind people based on an improved You Only Look Once (YOLO) algorithm and binocular vision. This method first improves the YOLOv8 algorithm by introducing an angle loss function, SPD convolution and a more efficient local convolution structure, and proposes an I-YOLOv8-BBOD model suitable for blind people’s travel. Then, based on this model, a blind travel intelligent obstacle avoidance platform is built. The confusion matrix analysis showed that the research model performed the best and had significant advantages in reducing false positives and false negatives. Its precision, recall and F1 score were all above 0.90, indicating the best overall performance. The ablation experiment showed that after the improvement of each module, the mAP and the average accuracy at the 50% IoU threshold increased by 13.19% and 16.00%, significantly improving the detection accuracy. In the example application, the successful obstacle avoidance rate of the proposed obstacle avoidance platform exceeded 90%, and the false alarm rate was below 1.3%. The traditional platform could reach up to 5.4%, which was better than the previous platform. This indicates that the research method can accurately detect obstacles and road conditions, ensuring the safety of visually impaired people’s travel.
The long-term coexistence of connected and automated vehicles (CAVs) and human-driven vehicles (HDVs) in mixed traffic flows will continue to significantly impact merging behaviours and operational efficiency in freeway on-ramp merging areas. To improve merging efficiency at freeway on-ramps, this paper proposes a cooperative merging model for mixed traffic. First, the merging points for ramp vehicles are determined based on vehicle interaction characteristics in the merging area. Then, for the dynamic decision-making problem of merging sequences, considering the differences in cooperative merging between CAVs and HDVs, the optimal merging sequence for both single-vehicle and multi-vehicle platoon merging is determined by minimising the global deviation between the actual travel time and the expected minimum travel time. Finally, according to the predetermined merging sequence, a bidirectional speed guidance strategy based on virtual vehicle mapping is designed for both mainline and ramp vehicles to achieve cooperative optimisation of the merging sequence and vehicle trajectories, enabling the smooth merging of ramp vehicles into the mainline. Simulations verify that the proposed method performs better than FIFO and effectively improves merging efficiency under different CAV penetration rates. Particularly in low-to-medium penetration scenarios, it achieves a 25% delay optimisation rate and 22% reduction in travel time, with the optimisation effect improving to varying degrees as penetration rates increase. This establishes a foundation for traffic analysis and cooperative control in freeway on-ramp merging areas under mixed traffic environments.
This work investigates intent recognition and trajectory prediction for multiple types of traffic participants at an unsignalized intersection within connected intelligence environments based on bidirectional spatiotemporal attention network (Bi-STANet). An unsignalized intersection is used as the studied object, where the participants include Connected and Automated Vehicles (CAVs), Human Vehicles (HVs), bicyclists, and pedestrians. A novel method is proposed based on Bi-STANet. First, a multimodal spatiotemporal feature extraction model is constructed based on (2+1)-dimensional CNN ((2+1)D CNN), where grid encoding method is used to unify the spatial structure and 2D convolution is used to extract spatial features for capturing the disordered characteristics of participants. Temporal dynamics are modelled via 1D convolution along the time axis, enabling spatiotemporal decoupling. Second, a bidirectional dynamic interaction model is developed by integrating LSTM-based temporal feature extraction with (2+1)D CNN layers, where heterogeneous modality fusion is implemented through a Bidirectional Contextual Block (BiCoBlock). Finally, a model integrating dynamic interaction, intent recognition, and trajectory prediction is developed. The proposed method is validated through the inD dataset innovatively. The results show that the average accuracy of intent recognition can reach to 95.4%. Within a 3-second horizon, Average Displacement Error (ADE) and Final Displacement Error (FDE) can be reduced to 0.51 m and 0.64 m, respectively, compared with the best baseline model. In Ablation studies, intent recognition F1-score can be enhanced by 7.2%, and ADE and FDE of trajectory prediction can be enhanced by 41.4% and 39.0%, respectively.
