Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MTDT: A Multi-Task Deep Learning Digital Twin

Nooshin Yousefzadeh, Rahul Sengupta, Yashaswi Karnati, Anand Rangarajan, Sanjay Ranka

TL;DR

MTDT addresses the challenge of estimating multiple Measures of Effectiveness (MOEs) at urban intersections when high-resolution detector data are sparse. It integrates two Graph Attention Network (GAT) based primary modules for lane-wise exit and inflow waveform estimation with two CNN-based secondary modules for queue length and travel-time MOEs, all trained in a multi-task framework. The model is trained on large-scale SUMO simulations guided by real ATSPM data, and it demonstrates robust generalization to arbitrary intersection topologies while enabling GPU-accelerated, near-parallel computation. The work contributes a scalable predictive framework for data-driven traffic control, capable of scenario generation and impact analysis for signal timing and topology changes. Altogether, MTDT offers a pathway toward practical digital twins for smart intersections that can inform timing optimization and safety analyses.

Abstract

Traffic congestion has significant impacts on both the economy and the environment. Measures of Effectiveness (MOEs) have long been the standard for evaluating traffic intersections' level of service and operational efficiency. However, the scarcity of traditional high-resolution loop detector data (ATSPM) presents challenges in accurately measuring MOEs or capturing the intricate spatiotemporal characteristics inherent in urban intersection traffic. To address this challenge, we present a comprehensive intersection traffic flow simulation that utilizes a multi-task learning paradigm. This approach combines graph convolutions for primary estimating lane-wise exit and inflow with time series convolutions for secondary assessing multi-directional queue lengths and travel time distribution through any arbitrary urban traffic intersection. Compared to existing deep learning methodologies, the proposed Multi-Task Deep Learning Digital Twin (MTDT) distinguishes itself through its adaptability to local temporal and spatial features, such as signal timing plans, intersection topology, driving behaviors, and turning movement counts. We also show the benefit of multi-task learning in the effectiveness of individual traffic simulation tasks. Furthermore, our approach facilitates sequential computation and provides complete parallelization through GPU implementation. This not only streamlines the computational process but also enhances scalability and performance.

MTDT: A Multi-Task Deep Learning Digital Twin

TL;DR

MTDT addresses the challenge of estimating multiple Measures of Effectiveness (MOEs) at urban intersections when high-resolution detector data are sparse. It integrates two Graph Attention Network (GAT) based primary modules for lane-wise exit and inflow waveform estimation with two CNN-based secondary modules for queue length and travel-time MOEs, all trained in a multi-task framework. The model is trained on large-scale SUMO simulations guided by real ATSPM data, and it demonstrates robust generalization to arbitrary intersection topologies while enabling GPU-accelerated, near-parallel computation. The work contributes a scalable predictive framework for data-driven traffic control, capable of scenario generation and impact analysis for signal timing and topology changes. Altogether, MTDT offers a pathway toward practical digital twins for smart intersections that can inform timing optimization and safety analyses.

Abstract

Traffic congestion has significant impacts on both the economy and the environment. Measures of Effectiveness (MOEs) have long been the standard for evaluating traffic intersections' level of service and operational efficiency. However, the scarcity of traditional high-resolution loop detector data (ATSPM) presents challenges in accurately measuring MOEs or capturing the intricate spatiotemporal characteristics inherent in urban intersection traffic. To address this challenge, we present a comprehensive intersection traffic flow simulation that utilizes a multi-task learning paradigm. This approach combines graph convolutions for primary estimating lane-wise exit and inflow with time series convolutions for secondary assessing multi-directional queue lengths and travel time distribution through any arbitrary urban traffic intersection. Compared to existing deep learning methodologies, the proposed Multi-Task Deep Learning Digital Twin (MTDT) distinguishes itself through its adaptability to local temporal and spatial features, such as signal timing plans, intersection topology, driving behaviors, and turning movement counts. We also show the benefit of multi-task learning in the effectiveness of individual traffic simulation tasks. Furthermore, our approach facilitates sequential computation and provides complete parallelization through GPU implementation. This not only streamlines the computational process but also enhances scalability and performance.
Paper Structure (13 sections, 3 equations, 2 figures, 4 tables)

This paper contains 13 sections, 3 equations, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Problem Definition
  3. Proposed Models
  4. Exit Module
  5. Inflow Module
  6. Queue Length Module
  7. Travel Time Module
  8. Loss Construction
  9. Traffic Data Generation
  10. Experimental Results
  11. Related Work
  12. Conclusions and Future Work
  13. Acknowledgments

Figures (2)

  • Figure 1: The inputs and outputs of Deep Multi-Task Learning Digital Twin We simulate the ATSPM time series waveform along with waveform or histogram of several MOEs within 8-phase standard NEMA phasing intersections. MTDT simultaneously estimates downstream exit waveforms of every outflow lane in all directions and upstream inflow waveforms of every inflow lane in all directions with an estimation of instant maximum queue length and travel time distribution associated with each phase of movement for an intersection with arbitrary topology and traffic conditions.
  • Figure 2: Overview of the architecture of MTDT. The architecture of the proposed Multi-componential digital twin framework contains two types of modules. GAT modules (light orange colored) execute primary tasks, which aggregated outputs are used as input to secondary tasks executed by CNN modules (dark blue colored). Red-colored arrows are activated only in inference mode.