Description

Urban transportation networks are increasingly affected by everyday congestion, non-recurring disruptions, and limited real-time visibility into what is happening on the road. Transportation agencies require advanced tools that support faster, proactive management and data-driven infrastructure decisions rather than passive monitoring. This report presents SMART-TWIN, an Artificial Intelligence (AI)–driven Digital Twin framework designed for real-time traffic operations, disruption detection, and performance-based decision-making support. The SMART-TWIN architecture integrates microscopic traffic simulation (detailed traffic modeling), long-horizon synthetic demand generation simulating real travel patterns, supervised machine learning, and performance analysis within a unified and reproducible workflow. A corridor-scale Digital Twin was developed for the Shaw–Cedar intersection in Fresno, California, using OpenStreetMap data and the SUMO simulation platform. An 11-year synthetic traffic dataset (2020–2030) was generated to emulate realistic traffic patterns over time and support early-stage deployment prior to full sensor availability. Supervised machine learning models were trained to classify operational conditions, including normal traffic, signal failure–induced stop-and-go conditions, and one- and two-lane closures. A Random Forest classifier achieved very high accuracy under controlled experimental conditions. To support operational decision-making, the framework translates detected traffic states into clear, measurable impacts, including estimated congestion buildup and delay-related costs using value-of-time principles. Scenario and sensitivity analyses demonstrate the compounding effects of capacity loss and control failure on congestion growth and user delay, as well as the system’s ability to detect disruptions early and reliably under moderate demand uncertainty. The results show that AI-enhanced Digital Twins can provide actionable insights into traffic conditions and operational impacts prior to full-field deployment. The proposed framework is scalable to larger corridors and networks and is designed for future integration with real-time sensor data and adaptive traffic control strategies. SMART-TWIN provides a practical foundation for next-generation Digital Twin applications that support proactive traffic management and data-driven infrastructure planning to make traveling safer and more efficient for all.

Publication Date

7-2026

Publication Type

Report

Topic

Transportation Technology, Transportation Engineering

Digital Object Identifier

10.31979/mti.2026.2537

MTI Project

2537

Mineta Transportation Institute URL

https://transweb.sjsu.edu/research/2537-Modeling-Real-Time-Transportation-Digital-Twins

Keywords

Digital twin, Intelligent transportation systems (ITS), Traffic Operations, Microscopic traffic simulation, Sumo, Machine learning, Incident detection, Traffic disruption analysis

Disciplines

Transportation | Urban Studies and Planning

Download

Additional files available below

Share

COinS