Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Heterogeneous Graph Sequence Neural Networks for Dynamic Traffic Assignment

Tong Liu, Hadi Meidani

TL;DR

This work tackles the problem of predicting link-level traffic flows across an entire urban network in scenarios with incomplete sensor coverage. It introduces HSTGSN, a heterogeneous spatio-temporal graph sequence network that jointly models road links and origin-destination (OD) links through an adaptive attention-based encoder-decoder to map OD demands to flow distributions under dynamic user equilibrium. The approach can operate with complete or incomplete OD information and can impute missing OD data via learned graph embeddings, achieving superior accuracy and generalization on three real-world networks compared to strong baselines. The results demonstrate the method's practical impact for scalable, sensor-agnostic traffic forecasting and dynamic traffic assignment in urban planning and management contexts.

Abstract

Traffic assignment and traffic flow prediction provide critical insights for urban planning, traffic management, and the development of intelligent transportation systems. An efficient model for calculating traffic flows over the entire transportation network could provide a more detailed and realistic understanding of traffic dynamics. However, existing traffic prediction approaches, such as those utilizing graph neural networks, are typically limited to locations where sensors are deployed and cannot predict traffic flows beyond sensor locations. To alleviate this limitation, inspired by fundamental relationship that exists between link flows and the origin-destination (OD) travel demands, we proposed the Heterogeneous Spatio-Temporal Graph Sequence Network (HSTGSN). HSTGSN exploits dependency between origin and destination nodes, even when it is long-range, and learns implicit vehicle route choices under different origin-destination demands. This model is based on a heterogeneous graph which consists of road links, OD links (virtual links connecting origins and destinations) and a spatio-temporal graph encoder-decoder that captures the spatio-temporal relationship between OD demands and flow distribution. We will show how the graph encoder-decoder is able to recover the incomplete information in the OD demand, by using node embedding from the graph decoder to predict the temporal changes in flow distribution. Using extensive experimental studies on real-world networks with complete/incomplete OD demands, we demonstrate that our method can not only capture the implicit spatio-temporal relationship between link traffic flows and OD demands but also achieve accurate prediction performance and generalization capability.

Heterogeneous Graph Sequence Neural Networks for Dynamic Traffic Assignment

TL;DR

This work tackles the problem of predicting link-level traffic flows across an entire urban network in scenarios with incomplete sensor coverage. It introduces HSTGSN, a heterogeneous spatio-temporal graph sequence network that jointly models road links and origin-destination (OD) links through an adaptive attention-based encoder-decoder to map OD demands to flow distributions under dynamic user equilibrium. The approach can operate with complete or incomplete OD information and can impute missing OD data via learned graph embeddings, achieving superior accuracy and generalization on three real-world networks compared to strong baselines. The results demonstrate the method's practical impact for scalable, sensor-agnostic traffic forecasting and dynamic traffic assignment in urban planning and management contexts.

Abstract

Traffic assignment and traffic flow prediction provide critical insights for urban planning, traffic management, and the development of intelligent transportation systems. An efficient model for calculating traffic flows over the entire transportation network could provide a more detailed and realistic understanding of traffic dynamics. However, existing traffic prediction approaches, such as those utilizing graph neural networks, are typically limited to locations where sensors are deployed and cannot predict traffic flows beyond sensor locations. To alleviate this limitation, inspired by fundamental relationship that exists between link flows and the origin-destination (OD) travel demands, we proposed the Heterogeneous Spatio-Temporal Graph Sequence Network (HSTGSN). HSTGSN exploits dependency between origin and destination nodes, even when it is long-range, and learns implicit vehicle route choices under different origin-destination demands. This model is based on a heterogeneous graph which consists of road links, OD links (virtual links connecting origins and destinations) and a spatio-temporal graph encoder-decoder that captures the spatio-temporal relationship between OD demands and flow distribution. We will show how the graph encoder-decoder is able to recover the incomplete information in the OD demand, by using node embedding from the graph decoder to predict the temporal changes in flow distribution. Using extensive experimental studies on real-world networks with complete/incomplete OD demands, we demonstrate that our method can not only capture the implicit spatio-temporal relationship between link traffic flows and OD demands but also achieve accurate prediction performance and generalization capability.
Paper Structure (21 sections, 12 equations, 5 figures, 3 tables)

This paper contains 21 sections, 12 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Dynamic User Equalibrium
  4. DUE-based Dynamic Traffic Assignment
  5. Proposed Method
  6. Heterogeneous Graph Representation
  7. Spatio-Temporal Graph Encoder
  8. Spatial Virtual Graph Encoder (S-VGE)
  9. Spatial Real Graph Encoder (S-RGE)
  10. Temporal Real-Graph Encoder (T-RGE)
  11. Spatio-Temporal Graph Decoder
  12. Link-level Traffic Flow Prediction
  13. Experiments
  14. Experiment Setup
  15. Dataset
  16. ...and 6 more sections

Figures (5)

  • Figure 1: Illustration of road links (solid lines) and OD links (dash lines). The information exchange between Origin and Destination on road links require two message passing steps (shown by the red curved arrows). When the OD link is added, this exchange is achieved after only one message passing step.
  • Figure 2: Illustration of the heterogeneous spatio-temporal graph sequence network. The solid and dashed lines in the graphs show road and OD links, respectively. The main component of HSTGSN is the spatio-temporal encoder-decoder. It includes five distinct components: the spatial virtual graph encoder (S-VGE), the spatial real graph encoder (S-RGE), and the temporal real graph encoder (T-RGE), the spatial real graph decoder (S-RGD), and the temporal real graph decoder (T-RGD).
  • Figure 3: Illustration of spatial-temporal graph encoder
  • Figure 4: Illustration of spatial-temporal graph decoder
  • Figure 5: Ablation study of different variants of HSTGSN. Three variations of HSTGSN are included: "w/o link feat": model maintaining OD link but removing road link features; "w/o OD link": model removing OD link; "w/o Adaptive": model removing adaptive attention mechanism.