Qiayuan Liao
Qiayuan Liao pursued doctoral research at Berkeley with Koushil Sreenath and Pieter Abbeel and earlier interned at Boston Dynamics on reinforcement-learning control. His work covers mechanisms and learning-based locomotion and manipulation. He co-authored Berkeley Humanoid, a 2024 research platform that used reinforcement learning for outdoor locomotion and hopping. He also co-authored MuJoCo Playground and its 2025 RSS demonstration. His work includes open-source control software and robot platforms.
- Works
- 24
- Citations
- 125
- h-index
- 7
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
Earlier career
Learning and physical robot systems
Liao pursued doctoral research at Berkeley and earlier interned at Boston Dynamics on reinforcement-learning control for E-Atlas. His research combines mechanism design with learning, planning, estimation, and control.
-
2024
Berkeley Humanoid
Liao co-authored Berkeley Humanoid in 2024. The platform was designed for learning-based control with low simulation complexity. The authors report using reinforcement learning with light domain randomisation to transfer locomotion policies from simulation to their physical robot.
-
2024
Testing physical locomotion
The platform demonstrated walking over several terrains and single- and double-leg hopping.
-
2025
A shared robot learning framework
Liao co-authored Demonstrating MuJoCo Playground at RSS 2025. The framework integrates simulation, rendering and training for several robot types. His open-source software covers whole-body tracking and legged robot control.
Sources and credits
Photo: qiayuanl.github.io; Qiayuan Liao; creator not stated. License: Not stated by source. Source: qiayuanl.github.io. Cropped to a square and resized. Ask us to remove or replace this photo.