Whole Body Control for Humanoid Robot Loco-Manipulation
Avideh Zakhor, Professor
Electrical Engineering and Computer Science
Applications for Fall 2026 are closed for this project.
Humanoid locomotion and manipulation are usually done separately. However, there are many situations in the real world, where the humanoid has to do loco-manipulation, i.e. manipulate while it is locomoting. In this project, we will develop methods to enable a Unitree G1 robot to walk, run, and perform whole body loco-manipulation tasks involving human with object interaction. Some of the techniques we will investigate include Reinforcement Learning.
Role: * Read the literature on humanoid autonomy;
* Come up with innovate algorithms to achieve human body control using RL or alternatives
* Implement the algorithm in simulation to verify performance
* Run it on the actual robot ( sim to real transfer) to verify performance on the robot.
Qualifications: * Must have taken the RL course or gone through the entire CS 285 online course.
* Familiar with deep learning frameworks such as Pytorch.
* Experience with actual hardware is a plus.
* Extensive programming and software development experience.
* Preference will be given to students who can continue the project over the spring semester.
Hours: 12 or more hours
Related website: https://www-video.eecs.berkeley.edu/papers/tomsonqu2/IROS_2024___march_revision_final.pdf
Related website: https://www-video.eecs.berkeley.edu/papers/zixianzang/hexapod_IROS_2023_final.pdf