KAIST DRCD Lab
About
About KAIST DRCD Lab
KAIST DRCD Lab is the Dynamic Robot Control and Design Laboratory at the Korea Advanced Institute of Science and Technology in Daejeon. Led by Hae-Won Park, it develops legged robots and the hardware and control methods needed for dynamic movement. Its research combines optimisation-based control, model predictive control, actuator design and robot learning. The laboratory builds experimental platforms including KAIST Humanoid, the HOUND quadrupeds and the MARVEL climbing robot.
The lab uses these systems to investigate locomotion and contact with objects and terrain. KAIST Humanoid v0.9 combines the lab’s whole-body platform with arms developed by RAILAB. In its October 2026 demonstration, onboard visual feedback supported pushing and lifting large objects, using skills developed through reinforcement learning and human demonstrations. These platforms support academic experiments and technical publications; the current humanoid results are research demonstrations.
Team
- Hae-Won Park - Associate Professor, Mechanical Engineering - Directs KAIST’s Humanoid Robot Research Center and researches dynamic legged systems.
Resources
Media library
- KAIST Humanoid v0.9 — whole-body pushing and lifting
video · KAIST DRCD Lab
Source - HOUND — agile perceptive multi-skill locomotion
video · KAIST DRCD Lab
Source
Timeline
Events
- 2026-10-19 - Demonstration - KAIST DRCD Lab demonstrates humanoid visual pushing and lifting
The demonstrated tasks push objects weighing up to 35 kg and lift objects up to 20 kg; these are separate experimental results. RAILAB developed the arms used with the DRCD humanoid platform. Interaction skills were developed with deep reinforcement learning and human demonstrations to arrange the letters KAIST.
Source - 2023-10-26 - Demonstration - KAIST DRCD Lab’s HOUND completes 100-metre sprint in 19.87 seconds
The test took place on October 26, 2023, and KAIST announced the recognised record on December 15. A single controller trained in simulation through reinforcement learning drove the robot. The lab incorporated motor characteristics into simulation to exploit the actuator output within the robot’s physical limits.
Source
Relations
Relationships
- RAILAB, KAIST
Technology Partner - KAIST DRCD Lab and RAILAB collaborated on Humanoid v0.9; RAILAB developed the arms.
Context
Comparable profiles
- Toyota Motor Corporation
stack adjacency - Reinforcement Learning Control
- Fern
stack adjacency - Reinforcement Learning Control
- Eka Robotics
stack adjacency - Reinforcement Learning Control
Shared workflows
- Honda Robotics
workflow adjacency - Humanoid Robot Research
- X-Humanoid
workflow adjacency - Humanoid Robot Research
- Robotable
workflow adjacency - Humanoid Robot Research
Context tags
- Legged locomotion - Capability - Legged-robot model predictive control, learning and dynamic locomotion across HOUND, humanoid and climbing research platforms.
- Humanoid robot research - Workflow - Academic full-size humanoid research programme with custom actuators; no commercial product or operating-deployment status inferred.
- Reinforcement Learning Control - Stack Layer - Laboratory reports simulation pretraining and reinforcement learning transferred to physical HOUND hardware; a research control layer, not a measured commercial deployment.