Research

Three lines of work.

Three active research threads, each with a working thesis and a small body of representative papers. Linked papers are on the Publications page; code (as it's released) will live on Data & Code.

Generative Intelligence for Transportation Modeling

We develop deep generative models — diffusion, normalizing flows, GANs, and probabilistic mixtures — to address fundamental challenges in transportation data modeling. Work spans traffic state estimation from sparse observations, occupancy density estimation, trajectory generation, and probabilistic forecasting. Generative approaches let us move beyond deterministic point estimates and capture the inherent uncertainty and complex spatiotemporal correlations in transportation data.

Representative papers

Diffusion Normalizing flows Trajectories Imputation Forecasting

AI-Powered Connected and Automated Driving

We develop AI methods for connected and automated driving (CAV) and cooperative-ITS — spanning multimodal foundation and Vision-Language-Action (VLA) models that unify perception, language, and control; deep reinforcement learning for cooperative driving and vehicle control; and vision-language approaches for traffic surveillance such as vehicle identification and monitoring. The goal is safe, cooperative automation that interprets complex traffic scenes, reasons about driving scenarios, and works in real-world conditions while integrating with the surrounding transportation system.

Representative papers

Connected & automated vehicles VLA Cooperative driving Reinforcement learning Surveillance

TMC-Agent · LLM-Powered Traffic Management

We're building AI agents powered by large language models to assist Traffic Management Center (TMC) operations. TMC-Agent aims to augment human operators by automating the interpretation of traffic network data, enabling natural language interaction with traffic management systems, and supporting real-time decision-making. The work includes building benchmarks to evaluate LLM capability on traffic network files and exploring how foundation models apply to transportation operations and planning.

Representative papers

LLM agents TMC operations Benchmarks Bayesian optimization

Funding & support

MnDOT · LRRB · UMN CTS · UMN DSI · NRF Korea · Korea Transport Institute