Manipulator technical dossier · 01

SO-101: an open arm as a controlled research baseline

Decide what the public package can reproduce, what must be frozen in an as-built record, and what still requires physical validation.

Evidence statusOpen engineering reference · reproducible research baseline · production readiness not establishedSources checked 4 August 2026

Identify the object first

A familiar name cannot replace an exact version, configuration and evidence trail.

SO-101 is the revised follower/leader arm family documented in The Robot Studio’s SO-ARM100 repository and supported by LeRobot. It is a learning and data-collection platform—not a certified industrial manipulator.

Engineering asset map

State what the public package provides—then what still has to be verified.

01

Geometry

Publicly available

STEP and printable STL files for the arm family.

Verify before adoption

Confirm the exact SO-101 revision, arm role and gripper variant; STL presence does not establish tolerances or material performance.

02

BOM & assembly

Publicly available

Community-maintained part lists and guidance covering servos, printed parts, electronics and fasteners.

Verify before adoption

Freeze supplier, part number, voltage, firmware and substitutions in the as-built record.

03

Control & calibration

Publicly available

LeRobot drivers, named configurations, calibration and teleoperation workflows.

Verify before adoption

Calibration belongs to a specific mechanism and identifier; preserve offsets, ranges, software version and configuration.

04

Simulation & data

Publicly available

Simulation material plus dataset and policy interfaces through LeRobot.

Verify before adoption

Simulation does not establish payload, thermal limits, cable life, backlash or safe stopping on the physical build.

Layered licence reading

One open repository does not give the complete robot one uniform set of rights.

Asset layerKnown statusAdoption decision
Repository files

Top-level repository identifies Apache-2.0.

Retain notices and check subdirectories and linked assets for exceptions before redistribution.

Vendor components

Servos, cameras, electronics and documentation remain third-party products.

Open-source code does not grant vendor documentation, trademark, warranty or supply rights.

Datasets & policies

Terms attach to each dataset or model; they are not inherited from the framework or arm repository.

Record provenance, consent, personal-data status and model terms for every asset.

This table is a source-navigation and engineering-decision aid, not legal advice. Use the original licence text and qualified advice for an actual decision.

Reproducibility gates

Every gate leaves reviewable evidence. A stop condition cannot be crossed by assumption.

  1. 01

    Revision freeze

    Evidence to preserve

    Repository commit, CAD/BOM revision, arm role, material specification and substitutions.

    Stop condition

    Stop if geometry, firmware or actuator variants cannot be reconciled.

  2. 02

    Bench calibration

    Evidence to preserve

    Joint range, zero offsets, repeatability, current and temperature logs under bounded motion.

    Stop condition

    Stop on unstable calibration, unexpected motion, overheating or inconsistent end stops.

  3. 03

    Task validation

    Evidence to preserve

    Payload, gripper force, collision envelope, cable routing and stop behaviour for one declared task.

    Stop condition

    Do not expand the task or allow human contact beyond the tested envelope.

  4. 04

    Repeat build

    Evidence to preserve

    A second build follows the record and reproduces calibration and bounded task metrics.

    Stop condition

    A one-off successful demo is not reproducibility evidence.

FUURAA analysis

Turn sources into an actionable judgement—not a pile of links.

01

Best use

Low-cost manipulation education, teleoperation studies and controlled data collection where the build is part of the experiment.

02

Main risk

A popular BOM and successful tutorial can be mistaken for evidence of interchangeable parts or repeatable performance.

03

FUURAA judgement

Adopt it as a versioned reference. Engineering value begins when substitutions, calibration and failures are recorded—not when the first motion video works.

Applicability boundary

This dossier can support the next experiment; it cannot replace unfinished engineering.

Can support

  • A bounded desktop manipulation prototype
  • Teleoperation and demonstration-data experiments
  • Comparative calibration research

This page cannot replace

  • Verified mechanical drawings and materials
  • Electrical protection and product safety review
  • Reliability, compliance and manufacturing evidence
Continue into the robot product development pathFrom requirements and evidence to a bounded prototype →