KORTHOS / ECOSYSTEM / PRODUCT

AI Sapiens K1

About

About AI Sapiens K1

AI Sapiens K1 is ROBOTIS's bipedal humanoid platform for developing locomotion, upper-body motion and learned robot policies. It combines in-house DYNAMIXEL-Q actuators with onboard NVIDIA computing and a documented control stack, giving researchers a route from simulation and demonstrations to execution on physical hardware. Imitation-learning workflows use a leader-follower system to collect training examples, while reinforcement-learning tools support locomotion development.

K1 has 23 actuated joints across its arms, legs and waist. Its onboard software runs ROS2 Jazzy within a container on Cyclo OS, with impedance control connecting policy outputs to the actuator buses. ROBOTIS publishes software and development guides as part of an open-source programme. Mechanical bills of materials, STEP files and electrical design resources remain listed as forthcoming, so the public hardware documentation is still being expanded.

08 SEPT 2026

ROBOTIS Moves Beyond Actuators Toward a Full Humanoid Stack

01 SEPT 2026

Microduck Preorders Create a Servo Demand Signal

30 JUL 2026

Humanoid supply chain — 27–30 July 2026

15 JUN 2026

Korea's humanoid supplier story moves into the joint stack

Capabilities

Source-backed capabilities

  • Control modes: Teleoperation
  • Bipedal walking

    Claimed - Walk forwards, backwards and sideways and turn in response to controller gimbals.

    Source

Supply

Supply / Components

  • Built-in-house QDD actuator stack

    ROBOTIS says K0 is powered by Dynamixel Q QDD actuators built in-house for efficiency, low backlash, and torque-level control.

    Source

Research

  • ROBOTIS AI Sapiens K1 ROS 2 packages

    2026-07-21 - Research Publication - Official K1 humanoid development packages covering robot descriptions, bringup, controllers and simulation-to-real tools.

    Source
  • ROBOTIS Cyclo Lab simulation environments

    Research Publication - Cyclo Lab provides Isaac Lab environments, robot assets and reinforcement/imitation learning configurations for ROBOTIS robots. Exact configurations include FFW-SG2 pick-and-place and AI Sapiens K1 locomotion and motion imitation.

    Source

Context

Comparable Systems