Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Metalearning traffic assignment for network disruptions with graph convolutional neural networks

Serio Agriesti, Guido Cantelmo, Francisco Camara Pereira

TL;DR

This work combines a machine-learning model (graph convolutional neural network) with a meta-learning architecture to train the former to quickly adapt to new graph structures and demand patterns, so that it may easily be applied to scenarios in which changes in the road network and the demand happen simultaneously.

Abstract

Building machine-learning models for estimating traffic flows from OD matrices requires an appropriate design of the training process and a training dataset spanning over multiple regimes and dynamics. As machine-learning models rely heavily on historical data, their predictions are typically accurate only when future traffic patterns resemble those observed during training. However, their performance often degrades when there is a significant statistical discrepancy between historical and future conditions. This issue is particularly relevant in traffic forecasting when predictions are required for modified versions of the network, where the underlying graph structure changes due to events such as maintenance, public demonstrations, flooding, or other extreme disruptions. Ironically, these are precisely the situations in which reliable traffic predictions are most needed. In the presented work, we combine a machine-learning model (graph convolutional neural network) with a meta-learning architecture to train the former to quickly adapt to new graph structures and demand patterns, so that it may easily be applied to scenarios in which changes in the road network (the graph) and the demand (the node features) happen simultaneously. Our results show that the use of meta-learning allows the graph neural network to quickly adapt to unseen graphs (network closures) and OD matrixes while easing the burden of designing a training dataset that covers all relevant patterns for the practitioners. The proposed architecture achieves a R^2 of around 0.85 over unseen closures and OD matrixes.

Metalearning traffic assignment for network disruptions with graph convolutional neural networks

TL;DR

This work combines a machine-learning model (graph convolutional neural network) with a meta-learning architecture to train the former to quickly adapt to new graph structures and demand patterns, so that it may easily be applied to scenarios in which changes in the road network and the demand happen simultaneously.

Abstract

Building machine-learning models for estimating traffic flows from OD matrices requires an appropriate design of the training process and a training dataset spanning over multiple regimes and dynamics. As machine-learning models rely heavily on historical data, their predictions are typically accurate only when future traffic patterns resemble those observed during training. However, their performance often degrades when there is a significant statistical discrepancy between historical and future conditions. This issue is particularly relevant in traffic forecasting when predictions are required for modified versions of the network, where the underlying graph structure changes due to events such as maintenance, public demonstrations, flooding, or other extreme disruptions. Ironically, these are precisely the situations in which reliable traffic predictions are most needed. In the presented work, we combine a machine-learning model (graph convolutional neural network) with a meta-learning architecture to train the former to quickly adapt to new graph structures and demand patterns, so that it may easily be applied to scenarios in which changes in the road network (the graph) and the demand (the node features) happen simultaneously. Our results show that the use of meta-learning allows the graph neural network to quickly adapt to unseen graphs (network closures) and OD matrixes while easing the burden of designing a training dataset that covers all relevant patterns for the practitioners. The proposed architecture achieves a R^2 of around 0.85 over unseen closures and OD matrixes.
Paper Structure (15 sections, 4 equations, 4 figures, 2 tables, 1 algorithm)

This paper contains 15 sections, 4 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Meta-learning in transportation
  3. Meta-learning for traffic assignment
  4. Methodology
  5. The gatedGCN
  6. The MAML architecture
  7. The experimental setting
  8. The dataset
  9. Support and query set
  10. Test set
  11. Hyperparameters
  12. Results
  13. Relevance to traffic best practices
  14. Limitations and future research directions
  15. Conclusions

Figures (4)

  • Figure 1: Schematic representation of one training epoch for the gatedGCN. $OD$ and $E$ are matrixes with information about the OD matrix and the edges in the graph. $\hat{q}$ represents instead the estimated flows and $Loss' = SmoothL1Loss$b23
  • Figure 2: Meta-loss progression representing the improvements measured against the query set and in the outer loop, as the meta-training progresses
  • Figure 3: Scatter plots of predicted vs estimated flows. Each plot reports values for twice the number of links times the number of query OD matrices in the meta-test
  • Figure 4: Network variation in task 47 - the red arrows signal the links that have become directed (one direction has been closed) while the dashed links represent closures in both directions.