Title: Toward Robust, Contact-aware, and Autonomous Humanoid Loco-manipulation
Date: Monday, July 13, 2026
Time: 1pm ET
Location: Love 183 or Zoom
Zhaoyuan Gu
Robotics Ph.D. Candidate
George W. Woodruff School of Mechanical Engineering
Georgia Institute of Technology
Committee:
Dr. Ye Zhao, Advisor – George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology
Dr. Maegan Tucker – School of Electrical and Computer Engineering and George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology
Dr. Sehoon Ha – School of Interactive Computing, Georgia Institute of Technology
Dr. Patrick Wensing – Department of Aerospace and Mechanical Engineering, University of Notre Dame
Dr. Guanya Shi – Robotics Institute and School of Computer Science, Carnegie Mellon University
Abstract:
Humanoid robots are built to achieve human-level locomotion and manipulation in human-centered environments. However, deploying autonomous humanoid systems reliably in the real world remains challenging. Robots must maintain balance under disturbances, sense and regulate physical contact, and coordinate locomotion and manipulation for reliable task execution. To address this gap, this dissertation draws inspiration from human capabilities associated with the cerebellum, skin, and brain, and develops robotic counterparts toward robust and autonomous humanoid loco-manipulation.
First, we address the fundamental problem of locomotion stability under external disturbances. Inspired by the cerebellum’s role in balance, we develop formal-methods-based model predictive control to quantify locomotion robustness and improve disturbance recovery through online optimization. Second, to unlock contact-rich manipulation, we design WT-UMI, a whole-body tactile sensing system that captures distributed contact across large body areas, resembling the role of human skin. Finally, inspired by the brain’s role in coordination and adaptation, we leverage reinforcement learning fine-tuning to improve policy success rates and generalization beyond demonstrated capabilities. Together, these contributions advance a path toward humanoid robots that can walk robustly, manipulate through distributed body contact, and perform autonomous loco-manipulation in real-world applications.