I'm Hongyu Zhou. Currently, I am a Ph.D. candidate at the Dept. of Aerospace Engineering, University of Michigan, under the supervision of Prof. Vasileios Tzoumas. Prior to that, I received the B.Eng. degree in Naval Architecture & Ocean Engineering from Huazhong University of Science and Technology (HUST), and the M.Sc. degrees in Marine Technology from Norwegian University of Science and Technology (NTNU). During my master's study, I undertook research on intelligent transportation systems, advised by Prof. Roger Skjetne. I was a research intern at Intelligent Control Lab, the Robotics Institute, CMU, advised by Prof. Changliu Liu. I also remotely worked with Dr. Zhuo Xu and Dr. Jiachen Li on socially-aware navigation for mobile robots and group-aware dense crowd forecasting at Mechanical Systems Control Lab, UC Berkeley.
My research centers on developing adaptive, agile, and data-efficient robot control algorithms under uncertainty. Building on online model learning and domain-adaptive policy optimization, my work enables robots to efficiently learn and adapt to modeling errors, environmental disturbances, and unknown behaviors of surrounding agents. My ongoing research focuses on structured and efficient policy learning and data generation for long-horizon robotic tasks. I am particularly interested in how to efficiently enable robots to adapt to new environments and tasks, transfer skills across different robot embodiments, and accomplish long-horizon tasks using only local capabilities.
My research has been mainly applied but not limited to quadrotors, including reference trajectory tracking under unknown aerodynamic effects (e.g., drag, ground effects, and wind diturbances), tracking target whose future behaviors are unknown a priori (structured, random, adversarial, or mixing of structured and random/adversarial), and safe autonomous flight in cluttered environments.
* indicates equal contributions, † indicates alphabetical order
