A panoramic view of downtown Salt Lake City skyline and the Wasatch Mountains, as seen from the University of Utah.

Chenxi Liu Lab

The Chenxi Liu Lab at the University of Utah bridges AI technologies and practical challenges in transportation engineering. Working toward a symbiotic system of people, AI, and the physical world, we study cyber-physical cooperation, trustworthy AI, and human-AI interaction to make transportation safer, more efficient, and more resilient.

About the Lab

About Dr. Chenxi (Dylan) Liu

Chenxi (Dylan) Liu is an Assistant Professor in the Department of Civil and Environmental Engineering at the University of Utah, where he leads the Chenxi Liu Lab. His research bridges AI technologies and practical challenges in transportation engineering to enhance the safety, efficiency, and resilience of transportation systems. His long-term goal is a symbiotic system of people, AI, and the physical world, in which each informs and supports the others. Three research directions build toward this goal. Cyber-physical cooperation connects AI with the physical transportation system, so that AI can perceive and understand its environment, share knowledge across vehicles and infrastructure, and act in coordination with the real world. Trustworthy AI makes the AI at the center of this system dependable, verifiable, and transparent, so that trust in it rests on evidence. Human-AI interaction connects AI with the people it serves, so that AI understands human behavior, intent, and judgment, people understand AI’s reasoning and limits, and the two work as partners. The lab carries this work onto real roads with the Utah Department of Transportation, Utah cities, and tribal communities, with support from USDOT, NSF, the Transportation Research Board, and UDOT.

Portrait of Professor Chenxi Liu (Dylan), primary investigator at the University of Utah, wearing glasses and a professional suit.

Education

Ph.D., Civil and Environmental Engineering (Transportation)
University of Washington, Seattle, WA
Dissertation: Situation-aware Customized Machine Intelligence for Transportation Safety, Equity, and Resilience
2020 – 2024
M.S., Transportation Engineering
University of Washington, Seattle, WA
Thesis: Bi-level Optimization Algorithm for Dynamic Reversible Lane Control Based on Short-term Traffic Flow Prediction
2017 – 2020
B.S., Civil Engineering
Tsinghua University, Beijing, China
Thesis: The Impacts of E-hailing Taxis on Urban Transportation Network
2013 – 2017

Research Focus

REACT trajectory predictions compared with V2X-ViT, DiscoNet, and CoBEVT across eight accident scenarios

Figure from REACT (Engineering Applications of Artificial Intelligence, 2026)

Cyber-Physical Cooperation

AI in transportation has to be grounded in the physical system it serves. We study how AI perceives and understands the physical environment, how knowledge is shared across vehicles and infrastructure, and how that understanding becomes coordinated action in the real world.

Featured paper · Engineering Applications of Artificial Intelligence 2026

Edge-Based Multimodal Sensor Data Fusion with Vision-Language-Action (VLA) Model for Real-Time Autonomous Vehicle Accident Avoidance

REACT combines hazard alerts broadcast by roadside units with the vehicle’s own camera views, and a lightweight vision-language-action model running on edge hardware re-plans the trajectory when an accident risk appears. On the DeepAccident benchmark it reduces the collision rate by 84.9% with 0.57-second inference latency on an NVIDIA Jetson AGX Orin.

View full publication list →

Overview of IDM-GPT with input validation, self-optimization prompting, database interaction, data analysis, and self-supervision modules

Figure from IDM-GPT (Transportation Research Record, 2025)

Trustworthy AI

Capability alone does not earn AI a role in safety-critical transportation decisions. We study how to make AI decisions dependable, verifiable, and transparent, so that trust in AI rests on evidence of how it performs and why it acts.

Featured paper · Transportation Research Record 2025

A Self-Supervised Multi-Agent Large Language Model Framework for Customized Traffic Mobility Analysis Using Machine Learning Models

In IDM-GPT, separate LLM agents validate each user query, review the prompts, and check the analysis results before a report is released, and the framework builds in privacy preservation. Users without training in transportation or machine learning can obtain reliable, customized traffic analysis in near real time.

View full publication list →

A student driving in the lab driving simulator

Human-AI Interaction

Transportation is shared by people and AI, and each must understand the other to work well together. We study how AI can learn human behavior, intent, and judgment, and how people can understand AI’s reasoning and limits, so that the two form an effective partnership.

Featured project · NSF National AI Research Resource Pilot · 2026

INTERACT: Intent-Driven Human–Vehicle Interaction with Constraint-Aware Control

Uses 9,200 GPU hours to train vehicle-control models that act on a person’s stated intent while staying inside explicit safety constraints.

View full project list →

Interested in joining the Chenxi Liu Lab?

We welcome Ph.D. students, master’s students, postdoctoral researchers, visiting scholars, undergraduate researchers, and summer interns who want to work at the intersection of AI and transportation.

How to Apply →

Latest News

All news →

NCHRP IDEA Project Awarded

October 2026

Our research project “Proactive V2X System for Beyond-Line-of-Sight (BLOS) Hazard Prevention” has been approved for funding by the NCHRP Highway IDEA Program of the Transportation Research Board, with Dr. Chenxi Liu as PI and Dr. Xiaoyue Cathy Liu as Co-PI.

USDOT SS4A Grant Awarded with Park City

October 2026

Park City, in partnership with the University of Utah and industry partner AIWaysion, received a $2.4M USDOT Safe Streets and Roads for All (SS4A) Planning and Demonstration Grant for “AI-Enabled Crash-to-Care Integration for Park City.” The project will pilot an AI-enabled crash-to-care system on three Park City corridors, including multimodal sensing, conflict prediction and warnings, automated crash reporting, and emergency response data sharing.

NSF NAIRR Pilot 9,200 GPU Hours

September 2026

Dr. Chenxi Liu and Dr. Xiaoyue Cathy Liu received 9,200 hours of advanced GPU computing through the National Science Foundation (NSF)-led National Artificial Intelligence Research Resource (NAIRR) Pilot to develop safer, more responsive AI-enabled vehicle-control systems.

Fengze Yang holding the 2026 ITE Mountain District Outstanding Graduate Student Award

Fengze Yang Receives ITE Mountain District Award

September 2026

Ph.D. student Fengze Yang received the ITE Mountain District Outstanding Graduate Student Award.

Predicted trajectories from the REACT framework across eight crash scenarios

New Paper in Engineering Applications of AI

September 2026

Our new research paper “Edge-Based Multimodal Sensor Data Fusion with Vision-Language-Action (VLA) Model for Real-Time Autonomous Vehicle Accident Avoidance” is published in Engineering Applications of Artificial Intelligence.

Welcome, Qingchang Yang

August 2026

We welcome Qingchang Yang, who joined the Chenxi Liu Lab as a Ph.D. student.

Students working on laptops during a class session

New Course: CVEEN 6592 AI-Enabled Electrified Transportation

August 2026

Dr. Chenxi Liu began teaching CVEEN 6592 AI-Enabled Electrified Transportation, a new course in the Transportation Electrification Certificate Program offered jointly by the University of Utah and Utah State University.

Traffic sign image samples from the ATSDAS study

New Paper in IEEE Transactions on ITS

August 2026

Our new research paper “ATSDAS: A Vision-Based System for Automated Traffic Sign Management and Condition Assessment” is published in IEEE Transactions on Intelligent Transportation Systems.

Mobile laser scanning point cloud of a rural road and roadside elements

New Paper in ASCE Journal of Transportation Engineering

August 2026

Our new research paper “Enhancing Deep Learning–Based Segmentation for Obstructive Roadside Element Detection in Point Cloud Data from Mobile Laser Scanning in Rural Road Environments” is published in ASCE Journal of Transportation Engineering, Part A: Systems.