Xuegang Ban
University of Washington, WA·Engineering
Modern roads produce huge amounts of data. Vehicle trajectories, travel speeds, traffic signal states, and video from cars and roadside cameras all help make traffic safer and smoother.
- Fields
Engineering
Computer Science
Mathematics & Statistics
- Topics
intelligent transportation systems
traffic dynamics
traffic flow
- Grants
- 3 known grants, through Sep 2028
What they work on
- Adapt and scale privacy-preserving techniques so that traffic, trajectory, and naturalistic driving data can be shared securely under both centralized and distributed data-sharing models, and evaluate them on real-world transportation datasets.
- Build a web-based recommendation system that helps stakeholders choose privacy techniques for their datasets, plus audit and compliance tools based on formal privacy guarantees, informed by user studies.
- Design new deep learning structures and a coordinated learning framework that discover the governing equations of traffic dynamics from noisy data, including mixed traffic with connected and automated vehicles (CAVs).
- Develop metrics and methods for judging learned traffic models on accuracy, parsimony (simplicity), interpretability, and generalizability to new conditions.
- Formulate data poisoning attacks on traffic state estimation and prediction as sensitivity analysis of optimization problems, using Lipschitz continuity analysis and semi-derivative algorithms, and test whether attacks transfer across deep learning models.
- Build an infrastructure-enabled defense framework that uses trusted, secure roadside and infrastructure data to detect and mitigate data poisoning attacks on transportation applications.
More about the research
These data also help in building self-driving cars. Relying on data brings new problems, though. The data can reveal private details about individuals, which makes it hard to share. It can be deliberately corrupted by attackers. And it is still not well understood how traffic will behave once automated vehicles share the road with human drivers. Across three projects, the lab tackles these problems: protecting privacy when traffic data is shared, learning the rules of traffic flow directly from data, and defending traffic prediction systems from data attacks.
Why it matters
Transportation agencies, automakers, and autonomous-vehicle developers could share driving data more safely while protecting people's privacy. Traffic models learned from data could help planners and traffic managers prepare for roads shared by automated and human-driven cars. Defenses against data poisoning could protect traffic prediction and control systems from manipulation, supporting safer, less congested, and more resilient transportation.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
PDaSP Track 2: A Holistic Privacy Preserving Collaborative Data Sharing System for Intelligent Transportation
NSF · Oct 2025 – Sep 2028
Collaborative Research: Towards Automatic Learning of Traffic Dynamics
NSF · May 2025 – Apr 2028
Collaborative Research: Data Poisoning Attacks and Infrastructure-Enabled Solutions for Traffic State Estimation and Prediction
NSF · Sep 2023 – Aug 2027
Closest labs
Michael Levin
University of Minnesota, MN·Engineering
Develop exact solution methods for transportation network design under user equilibrium that combine high-point relaxations, value-function cuts and strong duality cuts to find provably optimal designs faster.
Guang Wang
Florida State University, FL·Computer Science
Build cross-domain data curation tools for MobilityNet that combine real mobility datasets from different sources to reduce data bias.
Ziran Wang
Purdue University, IN·Engineering·Early-career
Determine how visual attention, spoken language and physical actions reveal a vehicle occupant's cognitive state during interaction with automation, using behavioral sensing.
The school
University of Washington: 81 labs on PI Pages, $1.7B a year on research, 694 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.