Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FusionTransNet for Smart Urban Mobility: Spatiotemporal Traffic Forecasting Through Multimodal Network Integration

Binwu Wang, Yan Leng, Guang Wang, Yang Wang

TL;DR

FusionTransNet tackles multimodal origin-destination traffic forecasting by integrating intra-modal spatiotemporal learning with a dual fusion inter-modal framework and a predictive decoder. It introduces OD-Adaptive-GCN with interconnected spatiotemporal graphs, Temporal Attention, Global/Local Fusion, ModeDistinctNet, and a multiple perspective interaction module, followed by an LSTM-based predictor. Empirical results on Shenzhen and New York demonstrate superior accuracy over diverse baselines and robust ablations validate each component’s value. The approach reveals cross-modal dependencies and provides a scalable framework transferable to other complex spatial systems such as supply chains or epidemic spread. The work advances multimodal urban mobility modeling and offers practical insights for planning, operations, and policy.

Abstract

This study develops FusionTransNet, a framework designed for Origin-Destination (OD) flow predictions within smart and multimodal urban transportation systems. Urban transportation complexity arises from the spatiotemporal interactions among various traffic modes. Motivated by analyzing multimodal data from Shenzhen, a framework that can dissect complicated spatiotemporal interactions between these modes, from the microscopic local level to the macroscopic city-wide perspective, is essential. The framework contains three core components: the Intra-modal Learning Module, the Inter-modal Learning Module, and the Prediction Decoder. The Intra-modal Learning Module is designed to analyze spatial dependencies within individual transportation modes, facilitating a granular understanding of single-mode spatiotemporal dynamics. The Inter-modal Learning Module extends this analysis, integrating data across different modes to uncover cross-modal interdependencies, by breaking down the interactions at both local and global scales. Finally, the Prediction Decoder synthesizes insights from the preceding modules to generate accurate OD flow predictions, translating complex multimodal interactions into forecasts. Empirical evaluations conducted in metropolitan contexts, including Shenzhen and New York, demonstrate FusionTransNet's superior predictive accuracy compared to existing state-of-the-art methods. The implication of this study extends beyond urban transportation, as the method for transferring information across different spatiotemporal graphs at both local and global scales can be instrumental in other spatial systems, such as supply chain logistics and epidemics spreading.

FusionTransNet for Smart Urban Mobility: Spatiotemporal Traffic Forecasting Through Multimodal Network Integration

TL;DR

FusionTransNet tackles multimodal origin-destination traffic forecasting by integrating intra-modal spatiotemporal learning with a dual fusion inter-modal framework and a predictive decoder. It introduces OD-Adaptive-GCN with interconnected spatiotemporal graphs, Temporal Attention, Global/Local Fusion, ModeDistinctNet, and a multiple perspective interaction module, followed by an LSTM-based predictor. Empirical results on Shenzhen and New York demonstrate superior accuracy over diverse baselines and robust ablations validate each component’s value. The approach reveals cross-modal dependencies and provides a scalable framework transferable to other complex spatial systems such as supply chains or epidemic spread. The work advances multimodal urban mobility modeling and offers practical insights for planning, operations, and policy.

Abstract

This study develops FusionTransNet, a framework designed for Origin-Destination (OD) flow predictions within smart and multimodal urban transportation systems. Urban transportation complexity arises from the spatiotemporal interactions among various traffic modes. Motivated by analyzing multimodal data from Shenzhen, a framework that can dissect complicated spatiotemporal interactions between these modes, from the microscopic local level to the macroscopic city-wide perspective, is essential. The framework contains three core components: the Intra-modal Learning Module, the Inter-modal Learning Module, and the Prediction Decoder. The Intra-modal Learning Module is designed to analyze spatial dependencies within individual transportation modes, facilitating a granular understanding of single-mode spatiotemporal dynamics. The Inter-modal Learning Module extends this analysis, integrating data across different modes to uncover cross-modal interdependencies, by breaking down the interactions at both local and global scales. Finally, the Prediction Decoder synthesizes insights from the preceding modules to generate accurate OD flow predictions, translating complex multimodal interactions into forecasts. Empirical evaluations conducted in metropolitan contexts, including Shenzhen and New York, demonstrate FusionTransNet's superior predictive accuracy compared to existing state-of-the-art methods. The implication of this study extends beyond urban transportation, as the method for transferring information across different spatiotemporal graphs at both local and global scales can be instrumental in other spatial systems, such as supply chain logistics and epidemics spreading.
Paper Structure (36 sections, 22 equations, 7 figures, 3 tables)

This paper contains 36 sections, 22 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Shallow Learning for OD Predictions
  4. Traditional Deep Learning for OD Prediction
  5. Graph Neural Networks for OD Prediction
  6. Model
  7. Problem Description
  8. Overview of the Proposed Learning Framework: FusionTransNet
  9. Phase 1 (Intra-Modal Learning Module).
  10. Phase 2 (Inter-Modal Learning Module):
  11. Phase 3 (Prediction Decoder Module):
  12. Phase 1: Intra-modal Learning in Urban Transportation Systems
  13. OD-Adaptive-GCN through Interconnected spatiotemporal Graphs.
  14. Temporal Attention Network.
  15. Phase 2: Inter-modal Learning: Dissecting Cross-Transportation-Mode Dynamics at Macro and Micro Levels
  16. ...and 21 more sections

Figures (7)

  • Figure 1: Motivating examples: Hourly traffic distribution of six regions in Shenzhen.
  • Figure 2: The structure of FusionTransNet, illustrating its components and goals.
  • Figure 3: Architecture of FusionTransNet, including the intra-modal learning phase, the inter-modal learning phase, and a final prediction decoder to predict the OD flow.
  • Figure 4: Ablation analysis of each component in the proposed framework. We abbreviate FusionTransNet as FTN on the $x$-axes. The first four bars represent ablation versions of our method, with the final bar indicating our complete method.
  • Figure 5: A case study of the global-fusion strategy (the red line shows the flow of the anchor).
  • ...and 2 more figures

Theorems & Definitions (4)

  • Definition 1: Origin-Destination graph
  • Example 1: Impact of rush hours on urban mobility
  • Example 2: Response to citywide events
  • Example 3: Necessity of Local Fusion