MMCAformer: Macro-Micro Cross-Attention Transformer for Traffic Speed Prediction with Microscopic Connected Vehicle Driving Behavior
arXiv:2602.16730v1 Announce Type: new Abstract: Accurate speed prediction is crucial for proactive traffic management to enhance traffic efficiency and safety. Existing studies have primarily relied on aggregated, macroscopic traffic flow data to predict future traffic trends, whereas road traffic dynamics are also influenced by individual, microscopic human driving behaviors. Recent Connected Vehicle (CV) data provide rich driving behavior features, offering new opportunities to incorporate these behavioral insights into speed prediction. To this end, we propose the Macro-Micro Cross-Attention Transformer (MMCAformer) to integrate CV data-based micro driving behavior features with macro traffic features for speed prediction. Specifically, MMCAformer employs self-attention to learn intrinsic dependencies in macro traffic flow and cross-attention to capture spatiotemporal interplays between macro traffic status and micro driving behavior. MMCAformer is optimized with a Student-t negati
arXiv:2602.16730v1 Announce Type: new Abstract: Accurate speed prediction is crucial for proactive traffic management to enhance traffic efficiency and safety. Existing studies have primarily relied on aggregated, macroscopic traffic flow data to predict future traffic trends, whereas road traffic dynamics are also influenced by individual, microscopic human driving behaviors. Recent Connected Vehicle (CV) data provide rich driving behavior features, offering new opportunities to incorporate these behavioral insights into speed prediction. To this end, we propose the Macro-Micro Cross-Attention Transformer (MMCAformer) to integrate CV data-based micro driving behavior features with macro traffic features for speed prediction. Specifically, MMCAformer employs self-attention to learn intrinsic dependencies in macro traffic flow and cross-attention to capture spatiotemporal interplays between macro traffic status and micro driving behavior. MMCAformer is optimized with a Student-t negative log-likelihood loss to provide point-wise speed prediction and estimate uncertainty. Experiments on four Florida freeways demonstrate the superior performance of the proposed MMCAformer compared to baselines. Compared with only using macro features, introducing micro driving behavior features not only enhances prediction accuracy (e.g., overall RMSE, MAE, and MAPE reduced by 9.0%, 6.9%, and 10.2%, respectively) but also shrinks model prediction uncertainty (e.g., mean predictive intervals decreased by 10.1-24.0% across the four freeways). Results reveal that hard braking and acceleration frequencies emerge as the most influential features. Such improvements are more pronounced under congested, low-speed traffic conditions.
Executive Summary
This research proposes the Macro-Micro Cross-Attention Transformer (MMCAformer) for traffic speed prediction, integrating macro traffic flow data with microscopic connected vehicle driving behavior features. Experiments on four Florida freeways demonstrate the superior performance of MMCAformer compared to baselines, with enhanced prediction accuracy and reduced model prediction uncertainty. The results reveal that hard braking and acceleration frequencies are the most influential features, particularly under congested traffic conditions. The MMCAformer model is optimized with a Student-t negative log-likelihood loss and provides point-wise speed prediction and uncertainty estimation. The findings have significant implications for proactive traffic management, enhancing traffic efficiency and safety. Furthermore, the study highlights the importance of considering both macro and micro traffic data in predicting traffic trends.
Key Points
- ▸ The MMCAformer model integrates macro traffic flow data with microscopic connected vehicle driving behavior features for traffic speed prediction.
- ▸ Experiments on four Florida freeways demonstrate the superior performance of MMCAformer compared to baselines.
- ▸ Hard braking and acceleration frequencies are the most influential features in traffic speed prediction, particularly under congested traffic conditions.
Merits
Strength in Handling Complex Traffic Dynamics
The MMCAformer model effectively captures the intrinsic dependencies in macro traffic flow and the spatiotemporal interplays between macro traffic status and micro driving behavior, enabling accurate traffic speed prediction under complex traffic conditions.
Improved Prediction Accuracy and Reduced Uncertainty
The MMCAformer model demonstrates superior performance compared to baselines, with enhanced prediction accuracy and reduced model prediction uncertainty, particularly under congested traffic conditions.
Incorporation of Microscopic Driving Behavior Features
The MMCAformer model incorporates microscopic connected vehicle driving behavior features, providing new insights into the influence of individual driving behaviors on traffic speed prediction.
Demerits
Limited Generalizability to Non-US Traffic Conditions
The study is limited to four Florida freeways, and the results may not be generalizable to non-US traffic conditions, which may have different traffic patterns and traffic management strategies.
Dependence on High-Quality CV Data
The MMCAformer model relies on high-quality connected vehicle data, which may not be readily available in all regions, limiting the practical applications of the model.
Computational Complexity and Training Requirements
The MMCAformer model is a complex neural network architecture, which may require significant computational resources and training time, limiting its practical applications in real-time traffic management systems.
Expert Commentary
The MMCAformer model is a significant contribution to the field of traffic speed prediction, demonstrating the potential of integrating macro and micro traffic data to improve prediction accuracy and reduce uncertainty. The study highlights the importance of considering both macro and micro traffic data in predicting traffic trends and managing traffic flow. The MMCAformer model has significant implications for urban mobility and traffic management, and its practical applications can be expanded to other regions and traffic management systems. However, the study is limited to four Florida freeways, and the results may not be generalizable to non-US traffic conditions. Additionally, the model relies on high-quality connected vehicle data, which may not be readily available in all regions. Nevertheless, the study demonstrates the potential of artificial intelligence and machine learning techniques in improving traffic speed prediction and traffic management, highlighting the need for further research in this area.
Recommendations
- ✓ Further research is needed to explore the generalizability of the MMCAformer model to non-US traffic conditions.
- ✓ The development of more efficient and scalable versions of the MMCAformer model is necessary to make it more practical for real-time traffic management systems.
