[论文解读] Reinforcement Learning-Guided Dynamic Multi-Graph Fusion for Evacuation Traffic Prediction

Md Nafees Fuad Rafi, Samiul Hasan|arXiv (Cornell University)|Jan 10, 2026

Traffic Prediction and Management Techniques被引用 0

一句话总结

介绍了 RL-DMF，一种将双动态图（距离与通行时间）与强化学习特征选择相结合的框架，用于预测飓风期间的疏散交通，达到最前沿的准确性和可解释性。

ABSTRACT

Real-time traffic prediction is critical for managing transportation systems during hurricane evacuations. Although data-driven graph-learning models have demonstrated strong capabilities in capturing the complex spatiotemporal dynamics of evacuation traffic at a network level, they mostly consider a single dimension (e.g., travel-time or distance) to construct the underlying graph. Furthermore, these models often lack interpretability, offering little insight into which input variables contribute most to their predictive performance. To overcome these limitations, we develop a novel Reinforcement Learning-guided Dynamic Multi-Graph Fusion (RL-DMF) framework for evacuation traffic prediction. We construct multiple dynamic graphs at each time step to represent heterogeneous spatiotemporal relationships between traffic detectors. A dynamic multi-graph fusion (DMF) module is employed to adaptively learn and combine information from these graphs. To enhance model interpretability, we introduce RL-based intelligent feature selection and ranking (RL-IFSR) method that learns to mask irrelevant features during model training. The model is evaluated using a real-world dataset of 12 hurricanes affecting Florida from 2016 to 2024. For an unseen hurricane (Milton, 2024), the model achieves a 95% accuracy (RMSE = 293.9) for predicting the next 1-hour traffic flow. Moreover, the model can forecast traffic flow for up to next 6 hours with 90% accuracy (RMSE = 426.4). The RL-DMF framework outperforms several state-of-the-art traffic prediction models. Furthermore, ablation experiments confirm the effectiveness of dynamic multi-graph fusion and RL-IFSR approaches for improving model performance. This research provides a generalized and interpretable model for real-time evacuation traffic forecasting, with significant implications for evacuation traffic management.

研究动机与目标

• 解决飓风中快速、随机的疏散交通模式，其中静态图和单特征模型表现不佳。
• 开发一个动态多图融合框架，联合捕捉基于距离的关系和基于通行时间的关系。
• 通过基于强化学习的智能特征选择与排序（RL-IFSR）提升模型可解释性。
• 在佛罗里达州的多个飓风和网络区段上展示泛化性与鲁棒性。

提出的方法

• 在每个时间步构建两个时变图：一个基于距离的图，另一个基于通行时间的图。
• 对每个图应用独立的图卷积，通过每个节点的注意力机制融合输出。
• 使用 LSTM 建模融合后节点嵌入的时序演化，并预测多步未来交通流量。
• 引入带有 Double Deep Q-Network 的 RL-IFSR，在训练过程中屏蔽低效特征并对特征重要性进行排序。
• 通过优先经验回放和 DDQN 目标来稳定学习并提高样本效率。

实验结果

研究问题

• RQ1 距离和通行时间图的动态多图融合是否能超越单图或静态图基线来改善疏散交通预测？
• RQ2 基于 RL 的特征屏蔽是否在数据稀缺和快速疏散动态下提升可解释性和鲁棒性？
• RQ3 所提出的框架在未见飓风和佛罗里达不同网络条件下的泛化能力如何？

主要发现

• RL-DMF 在评估指标上优于基线模型（LSTM、CNN-LSTM、静态 GCNN-LSTM、仅距离或仅通行时间图的动态 GCNN-LSTM）。
• 对于 Milton 飓风，1 小时预测的 RMSE 为 293.9，MAE 为 189.5，MAPE 为 17.9%，R2 为 0.95。
• 在 1–6 小时的预测区间内，RMSE 从 293.9 变动至 495.3，R2 始终高于 0.86，总体 RMSE 426.4，MAE 281.1，MAPE 25.2%，R2 0.90。
• DMF 通过利用距离图和通行时间图的互补空间依赖来提升预测。
• RL-IFSR 通过在训练过程中屏蔽低效特征，提供可解释的特征排序，在动态且数据稀缺的疏散场景中提升鲁棒性。