SUMMARY:

 

KAIST Humanoid v0.7 is a high‑speed bipedal locomotion research robot developed by the Dynamic Robot Control & Design (DRCD) Lab at KAIST. It demonstrates moonwalk, soccer dribbling, running, and agile whole‑body balance, powered by deep reinforcement learning and in‑house actuators.  

Sources: 

 

EDITORIAL:

 

KAIST’s Humanoid v0.7 is not a commercial humanoid — it is a locomotion‑first research platform designed to push the limits of agility, balance, and dynamic control. In a field crowded with acrobatic demos, v0.7 stands out because:

 

- It uses in‑house actuators and custom control hardware.  

- Its locomotion policies are trained using deep reinforcement learning with human demonstrations.  

- It performs moonwalk, soccer dribbling, high‑speed running, and robust balance even under disturbances.  

- It is part of KAIST’s long lineage of world‑class humanoids (KHR‑1 → HUBO → DRC‑HUBO → v0.5 → v0.7).

 

v0.7 is a pure research machine, built to test next‑generation locomotion algorithms that will eventually feed into industrial and disaster‑response humanoids.  

Sources: 

 

SPECIFICATIONS:

 

Dimensions & Build

(KAIST has not published exact height/weight for v0.7; only performance characteristics are public.)  

- Category: High‑speed bipedal locomotion research robot  

- Developer: KAIST DRCD Lab (Dynamic Robot Control & Design Laboratory)  

Sources: 

 

Mobility

- Locomotion: Bipedal walking, running, moonwalk, duck‑walk  

- Running speed: Up to 12 km/h (lower‑body platform benchmark)  

- Terrain robustness: Maintains balance even with eyes closed or on bumpy surfaces  

- Sports capability: Soccer dribbling & kicking  

Sources: 

 

Degrees of Freedom

(Not published; v0.7 is a research platform with custom DOF distribution.)  

- Multi‑DOF legs for dynamic locomotion  

- Upper‑body DOF sufficient for balance & interaction  

Sources: 

 

Actuation & Manipulation

- Actuators: In‑house KAIST electric actuators  

- Control: High‑bandwidth torque control for dynamic motions  

- Manipulation: Limited; platform is locomotion‑focused  

Sources: 

 

Perception & Sensors

- Vision: Cameras for soccer & interaction tasks  

- Inertial sensing: IMU for balance  

- Additional: Sensors for foot placement & ground reaction  

Sources: 

 

Compute & AI

- Control policy: Deep reinforcement learning  

- Training: Human demonstration + RL fine‑tuning  

- Behavior: Autonomous locomotion & interaction  

Sources: 

 

Battery & Power

- Not published (lab‑tethered or internal pack depending on test mode)

 

Connectivity

- Research‑grade control interface  

- Likely tethered + wireless for field tests (not published)

 

Applications

- High‑speed locomotion research  

- Reinforcement‑learning control studies  

- Sports‑skill robotics (soccer, dribbling)  

- Dynamic balance & agility research  

- Future industrial & disaster‑response locomotion foundations  

 

Pricing

- Not commercially available  

- Research‑only platform

 

WHAT KAIST HAS NOT PUBLISHED

- Height & weight  

- DOF breakdown  

- Battery capacity  

- Sensor resolution  

- Torque curves  

- Commercial roadmap  

 

Image: @TechnologyXAI

 

 

COMPANY SOURCE:

 

 

https://youtu.be/9qZcTMARvpk?si=i-n96f0_q9ujFacy

 

 

NEWS & KEY COVERAGE:

 

 

https://youtu.be/565JW8UhlrE?si=fyDP9qJEzyyL01EF