ATHENA SEMINAR SERIES: Customizing Trustworthy Machine Learning Methods for Smart and Equitable Cyber-physical Systems

Presenter: Hao (Frank) Yang

Abstract:

Civil cyber-physical systems (CPS) are intricate networks that combine physical infrastructure systems with digital technologies to provide public services. Fortunately, the proliferation of smart city sensors and personal mobile devices in recent years has generated vast amounts of valuable data that can aid in managing, forecasting, and controlling various civil CPSs. However, traditional information analysis techniques used by civil engineers are not adequate for handling big data. Although many general-purpose data-driven and artificial intelligence methods are available, they may not be appropriate for civil applications due to limitations such as insufficient and biased training data, lack of interpretability, limited scalability, and concerns over user equity and privacy. In this presentation, the author will discuss research efforts focused on customizing trustworthy machine learning methods for traffic systems. These methods will incorporate physical laws and traffic domain knowledge to enhance equity, safety, reliability, and sustainability in autonomous vehicles and large-scale traffic networks. The demonstrated benefits of these tailored machine learning methods in traffic systems can be easily extended to other civil cyber-physical systems as well. This highlights the promising potential of integrating domain knowledge from real-world systems into artificial intelligence research, leading to a brighter future of smarter and safer CPS technologies.

 

Bio:

Hao (Frank) Yang received B.S. degrees in Electrical and Computer (Telecommunication) Engineering from Beijing University of Posts and Telecommunications and the University of London in 2017. He is currently a postdoc/research scientist at Duke University, and the incoming tenure-track assistant professor at the Johns Hopkins University. His research focuses on developing trustworthy machine learning and data science methods to improve the equity, safety, resilience, and sustainability of traffic systems. Specifically, his research includes 1) the development of new sensors and perception methodologies, 2) the creation of physical-informed AI and machine learning methods, and 3) the building of human-machine cooperative traffic systems. He is an active member of three TRB standing committees and the student subcommittee chair of ASCE Transportation & Development Institute (T&DI) AI committee. He has served as an associated editor of the IEEE Intelligent Transportation Systems Conference since 2021. He is the recipient of the Michael Kyte Outstanding Student of the Year Award by US Department of Transportation for Federal Region 10 in 2022, the High-Value Research Award of American Association of State Highway and Transportation Officials (AASHTO) in 2022, and the Best Paper Award of the TRB Information Systems and Technology Committee (AED30) in 2023.