Learning Robustness and Generalizable Humanoid Loco-Manipulation from Human Demonstration

PhD Qualifying Examination


Title: "Learning Robustness and Generalizable Humanoid Loco-Manipulation from 
Human Demonstration"

by

Miss Runyi YU


Abstract:

Humanoid loco-manipulation remains a foundational challenge in robotics with 
significant implications for practical applications. While physical 
simulation and reinforcement learning (RL) have greatly advanced humanoid 
locomotion, enabling physical robots to perform agile and natural 
loco-manipulation remains a major challenge. This proposal investigates a 
promising paradigm centered on Reinforcement Learning from Interaction 
Demonstration (RLID), which involves capturing human-object interactions, 
retargeting them to a simulated humanoid robot, and leveraging imitation 
rewards to train robust and generalized policies with success sim-to-real 
transfer.


Date:                   Tuesday, 22 April 2026

Time:                   2:00pm - 4:00pm

Venue:                  Room 2132C
                        Lift 22

Committee Members:      Dr. Qifeng Chen (Supervisor)
                        Dr. Long Chen (Chairperson)
                        Prof. Chi-Keung Tang