Research

I embed traffic-flow theory into machine learning so that network states can be reconstructed where sensing is incomplete. This work includes observation-aligned two-stage domain decomposition for speed-field reconstruction from sparse fixed sensors, physics-informed formulations of macroscopic models such as the generalized bathtub model, and graph learning guided by the network macroscopic fundamental diagram for traffic state imputation. The aim is inference that scales to large networks and stays reliable where data are limited.

I study how transportation networks absorb disruption and how operations can be designed for faster recovery. This work quantifies how route guidance attacks and other disruptions propagate through a network, extends analytical network-flow models to traffic under manipulated guidance, and measures the network-level effects of operational changes such as movable lane management, connecting link-level intervention to network outcomes.

I model how travelers respond to guidance they cannot fully trust. This work includes user-centered study of behavioral sensitivity to route guidance attacks, day-to-day network models in which trust in guidance is endogenous and reliance adapts over time, and empirical driver-behavior studies using surrogate safety measures, linking individual response to aggregate network dynamics.

I model and mitigate cyber threats to connected and automated mobility. This work includes detection of V2X attacks such as fake emergency electronic brake light messages, bounded-rational route guidance attacks on urban navigation services, and a review of cybersecurity needs for next-generation road transportation, with defenses judged by their network-level outcomes rather than by detection accuracy alone.