arXiv：cs.LG· Donggyu Min, Dong-Kyu Kim·· 4 小时前AI 评分33
LFPG-RL：用强化学习与链路流量传播引导在线估计动态 OD 矩阵
Online Estimation of Dynamic Origin-Destination Matrices Using Reinforcement Learning with Link-Flow Propagation Guidance
AI 导读
研究提出 LFPG-RL，将模拟车辆传播记录与下游误差结合，引导动态 OD 矩阵各分量的在线估计。在墨尔本干路网 250 条工作日链路流量轨迹上，平均测试 RMSE 为 7.41 辆/15 分钟，MAPE 为 31.24%，Pearson 相关系数 0.987，较最强基线（带传播引导的梯度下降）RMSE 降低 40.1%。该方法可快速更新需求，使交通仿真与观测条件保持一致。
正文
Abstract:Dynamic origin-destination (OD) matrix estimation calibrates time-dependent input demand for simulations to reproduce observed link flows. Reinforcement learning is well suited to online estimation because a trained policy estimates demand in a single evaluation, but varying target flows complicate learning from aggregate rewards. We propose reinforcement learning with link-flow propagation guidance (LFPG-RL), combining simulated vehicle propagation records with downstream errors to guide individual OD components. Using 250 weekday link-flow trajectories from a Melbourne arterial network, LFPG-RL achieves a mean test root mean squared error (RMSE) of 7.41 vehicles per 15-minute interval, mean absolute percentage error of 31.24%, and Pearson correlation of 0.987. Comparisons with optimisation, filtering and reinforcement learning without guidance show a 40.1% RMSE reduction over the strongest baseline, gradient descent with propagation guidance. LFPG-RL enables rapid demand updates that keep traffic simulations consistent with observed conditions, supporting the evaluation of traffic management strategies.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.30317 [cs.LG] |
| (or arXiv:2608.30317v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2608.30317 arXiv-issued DOI via DataCite |
Submission history
From: Donggyu Min [view email]
[v1]
Mon, 31 Aug 2026 06:33:21 UTC (10,374 KB)
[v2]
Wed, 7 Oct 2026 07:38:42 UTC (11,461 KB)
来源：arXiv：cs.LG · arxiv.org