Background & experience

About me

I am Yukai Wang (王于凯), a PhD Candidate at KAIST’s Cho Chun Shik Graduate School of Mobility. My research combines traffic modeling, machine learning, and safety analysis, with a current focus on learning and evaluating multi-agent traffic world models from drone trajectory data.

I am advised by Prof. Tiantian Chen and work in the Human Factors Centered Transport Safety Laboratory. My earlier work at Southeast University, advised by Prof. Zhiyuan Liu, focused on transportation and learning-based decision methods.

My current direction grew from traffic-conflict and extreme-value research: I wanted to understand future interactions, not only measure their outcomes. Drone trajectories let me ask that question at the scale of a traffic site. My aim is to develop traffic world models for transportation research beyond a single autonomous vehicle. Read the research story and the connected questions that guide my ongoing work.

Education

Research experience

Doctoral research, KAIST · 2023–present. Multi-agent trajectory learning and traffic world models; closed-loop evaluation and rollout diagnostics; world-model reinforcement learning; traffic-conflict and extreme-value methods. See Research for project descriptions, figures, and current status.

Graduate research, Southeast University · 2020–2023. Transportation research with Prof. Zhiyuan Liu. Related collaborative work on deep reinforcement learning for joint travel-mode and departure-time choice appeared in Multimodal Transportation in 2024.

Industry experience

Selected awards

Research methods and tools

Graph neural networks; recurrent state-space models; neural ODEs; autoregressive trajectory models; reinforcement learning; surrogate safety measures; extreme value theory; recording-disjoint evaluation and bootstrap diagnostics. Implementation experience includes Python, PyTorch / PyTorch Lightning, CARLA-based simulation, Git, and LaTeX.