Skip to content

Tasks

Tasks are the core unit of work in Humanoid Network. Each task is a structured motion-generation brief designed to produce training data for a specific robotic skill.

Task Structure

Every task includes:

Field Description
Title Short description of the motion
ROBO Category Which robotic skill this trains
Prompt template Base text-prompt structure
Robot target Humanoid platform the motion must fit
Duration Expected motion length
Reward HAN Points earned for a validated submission
Difficulty Beginner, Intermediate, or Advanced

Task Types

Manipulation Tasks

Object interaction: picking, placing, stacking, pouring, opening, closing.

Movement through environments: walking, turning, obstacle avoidance, door traversal.

Social Interaction Tasks

Human-facing behaviors: gesturing, pointing, handing objects, following directions.

Dexterous Tasks

Fine motor skills: typing, writing, tool use, button pressing, knob turning.

Task Lifecycle

  1. Published: Task is available for contributors to claim
  2. In Progress: A contributor is actively generating and refining
  3. Submitted: Motion uploaded, awaiting Hydra-Track validation
  4. Validated: Physics validation passed, HAN Points credited
  5. Failed: Validation did not pass, contributor notified with feedback

Task Design

Tasks are designed by the Humanoid Network team and partner robotics labs. Each task targets a specific gap in available training data, ensuring the network produces maximally useful motion datasets.