Learning-Based Whole-Body Control for Humanoid Robots

PhD Qualifying Examination


Title: "Learning-Based Whole-Body Control for Humanoid Robots"

by

Mr. Zhi LIN


Abstract:

Humanoid robots recently received increased attention for their unique
potential to perform tasks in human-centric environments. However, their
redundancy in DoF and instability nature pose significant challenges for
control. Though traditional model-based approaches have been developed for
decades, due to the high computational demands of solving sequences of
optimization problems and the difficulty of modeling environmental
uncertainty, progress has been limited. Recently, with the development of
physics simulators, training policies on massive parallel simulated
environments using learning-based approaches have become feasible. This
makes learning-based approaches now dominate this area. This survey reviews
the current landscape of learning-based whole-body control (WBC) for
humanoids. This review follows the standard development pipeline: Data
Acquisition and Preprocessing, Policy Training, and Deployment, and
organizes recent works by their targeted stages and motivations. Lastly, we
discuss the open challenges that remain to be solved in this domain.


Date:                   Tuesday, 16 December 2025

Time:                   4:00pm - 6:00pm

Venue:                  Room 2128B
                        Lift 19

Committee Members:      Prof. Nevin Zhang (Supervisor)
                        Prof. Fangzhen Lin (Chairperson)
                        Prof. Ping Tan (ECE)