MTDT: A Multi-Task Deep Learning Digital Twin

2024-05-02

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC)

AI Summary

Traffic intersection performance evaluation—measured via Metrics of Effectiveness (MOEs)—is hindered by the sparsity and limited spatial coverage of conventional loop detector data, impeding accurate characterization of complex spatiotemporal dynamics. To address this, we propose a multi-task deep learning digital twin model tailored for urban intersections. The model introduces a novel hybrid architecture integrating Graph Convolutional Networks (GCNs) and one-dimensional Temporal Convolutional Networks (TCNs) to jointly capture localized spatiotemporal features—including signal timing plans, road network topology, and turning movement flows—in an adaptive manner. Through multi-task joint learning, it simultaneously predicts lane-level inflow/outflow volumes, multi-directional queue lengths, and travel time distributions with high accuracy. Implemented for GPU-accelerated full parallel inference, the model ensures real-time execution and scalability. Microscopic simulation validation demonstrates substantial improvements in MOE estimation accuracy, establishing a new paradigm for intelligent traffic signal control and digital twin-enabled intersection management.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Mixture of Experts (MoE)
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

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 multitask 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 multitask 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.

Problem

• Traffic congestion impacts economy and environment
• Scarcity of high-resolution data challenges traffic evaluation
• MTDT simulates intersection flow using multi-task learning

Innovation

• Multi-task learning combines graph and time convolutions
• Adapts to local spatiotemporal features like signal timing
• Enables GPU parallelization for scalable computation