Out of the Past: An AI-Enabled Pipeline for Traffic Simulation from Noisy, Multimodal Detector Data and Stakeholder Feedback
🤖 AI Summary
Existing data-driven traffic simulation methods rely on unrealistic heuristics and neglect data uncertainty and multimodality. To address these limitations, this paper proposes an end-to-end AI simulation pipeline. The method introduces a three-stage synergistic framework integrating computer vision (for vehicle detection and counting), combinatorial optimization (for path inference via integer programming), and large language models (for natural language–driven iterative calibration). It is the first to enable NL-feedback-guided dynamic simulation correction. Evaluated on the Strongsville road network in Ohio, the system faithfully reproduces fine-grained traffic flow patterns. Moreover, it demonstrates strong cross-city generalizability and low data dependency—requiring only minimal labeled inputs—while significantly enhancing both simulation fidelity and interpretability.
Application Category
📝 Abstract
How can a traffic simulation be designed to faithfully reflect real-world traffic conditions? Past data-driven approaches to traffic simulation in the literature have relied on unrealistic or suboptimal heuristics. They also fail to adequately account for the effects of uncertainty and multimodality in the data on simulation outcomes. In this work, we integrate advances in AI to construct a three-step, end-to-end pipeline for generating a traffic simulation from detector data: computer vision for vehicle counting from camera footage, combinatorial optimization for vehicle route generation from multimodal data, and large language models for iterative simulation refinement from natural language feedback. Using a road network from Strongsville, Ohio as a testbed, we demonstrate that our pipeline can accurately capture the city's traffic patterns in a granular simulation. Beyond Strongsville, our traffic simulation framework can be generalized to other municipalities with different levels of data and infrastructure availability.
Problem
Research questions and friction points this paper is trying to address.
Designing traffic simulations that reflect real-world conditions accurately
Addressing uncertainty and multimodality in traffic data for simulations
Creating a generalizable pipeline for municipalities with varying data availability
Innovation
Methods, ideas, or system contributions that make the work stand out.
Computer vision for vehicle counting
Combinatorial optimization for route generation
Large language models for feedback refinement
🔎 Similar Papers
No similar papers found.
