KORTHOS / ECOSYSTEM / PRODUCT

Linker Hand L20

About

About Linker Hand L20

Linker Hand L20 is a five-finger robotic hand built around brushless motors, ball-screw transmission and articulated linkages. Its 16 actively driven and five passive degrees of freedom provide coordinated finger motion, with self-developed joint modules supporting position and torque control. The mechanical design places actuation and transmission within a compact hand assembly.

Developers connect through CAN or RS485, and LinkerBot provides Windows and Ubuntu host interfaces for operation. A 24–48 V supply supports integration with robot hardware and industrial power systems. The hand is intended for manipulation development involving grasping, assembly and object handling, with independent finger articulation providing more control over contact geometry than a simple opening-and-closing grasp.

17 JUL 2026

China’s Dexterous-Hand Suppliers Are Becoming Unicorns

17 MAY 2026

It’s all in the hands

Capabilities

Source-backed capabilities

  • Capability areas: Robotic grasping
  • Workflows: Multi-finger grasping, tactile or sensor-rich manipulation, tool interaction and advanced end-effector research.
  • Operating environments: Humanoid platforms, robot-arm cells, research labs and embodied-AI manipulation environments.
  • Target markets: Humanoid builders, automation teams, researchers and developers needing more hand capability than simple grippers provide.
  • Deployment model: Commercial hand product in LinkerBot’s dexterous-hand lineup.
  • Robotic grasping

    Research demonstration - 2026-01-13 - The FSAG study transfers finger-specific grasp affordances from human video to physical Linker Hand L20 hardware without per-hand perception retraining, evaluating stable multi-contact grasps across seen and unseen bottles, fruit and tools.

    Source
  • Dexterous manipulation

    Research demonstration - 2026-01-13 - The FSAG study transfers finger-specific grasp affordances from human video to physical Linker Hand L20 hardware without per-hand perception retraining, evaluating stable multi-contact grasps across seen and unseen bottles, fruit and tools.

    Source

Research

  • DexNinja: Learning Robust Dexterous Cutting Policy with a Real-to-Sim-to-Real Data Engine

    2026-10-12 - Research Publication - DexNinja combines limited real demonstrations, reconstructed food assets and differentiable cutting simulation to train tactile dexterous policies. Experiments study food slicing and transfer across two bimanual arm-and-hand configurations.

    Source
  • UniDexTok: A Unified Dexterous Hand Tokenizer from Real Data

    2026-06-09 - Research Publication - UniDexTok learns a shared semantic state representation from real data across heterogeneous dexterous hands, including LinkerHand L10 and L20. The paper reports sub-millimetre reconstruction and strong zero- and few-shot transfer to new hand embodiments.

    Source
  • FSAG: Enhancing Human-to-Dexterous-Hand Finger-Specific Affordance Grounding via Diffusion Models

    2026-01-13 - Research Publication - FSAG transfers finger-specific semantic grasp targets from human video to physical dexterous hands without per-hand retraining. The paper evaluates real grasping on Linker Hand L20 and reports strong cross-hand and unseen-object generalization for bottles, fruit and tools.

    Source

Context

Market context

20-DoF multi-sensor dexterous robotic hand for human-like grasping and robotic manipulation.

  • Target environment: Humanoid platforms, robot-arm cells, research labs and embodied-AI manipulation environments.
  • Workflow context: Multi-finger grasping, tactile or sensor-rich manipulation, tool interaction and advanced end-effector research.
  • Customer context: Humanoid builders, automation teams, researchers and developers needing more hand capability than simple grippers provide.
  • Deployment model: Commercial hand product in LinkerBot’s dexterous-hand lineup.
  • Commercial maturity: Commercially positioned product with official family-level product materials.
  • Adoption constraints: Adoption depends on sensing needs, finger durability, control stack, mechanical fit, price and support for the target robot.
  • Market position: Mid-to-high capability dexterous-hand lane for builders needing more articulation than entry-level hands.
  • Adjacent products: Linker Hand L10, Linker Hand L30, AGILINK OmniHand, Shadow Dexterous Hand

Comparable Systems