Contact us

Manipulator technical dossier · 03

ALOHA: a bimanual system as a complete evidence chain

Trace the original open research stack from operator input and synchronized sensing to recorded demonstrations, learned policies and later platform generations.

Evidence statusInfluential academic reference · configuration-specific stack · later variants require separate verificationSources checked 4 August 2026
Read the engineering asset map ↓Back to arm references ↗

Identify the object first

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

ALOHA is an influential low-cost bimanual teleoperation and data-collection research system. The original repository, ACT work and later ALOHA 2 / Mobile ALOHA research are related but not interchangeable packages.

Engineering asset map

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

01

Teleoperation

Publicly available

Leader/follower configuration, joint control, camera setup and coordinated bimanual operation.

Verify before adoption

Match every arm, actuator, firmware, port and control mode to the actual configuration.

02

Data collection

Publicly available

Episode recording and replay with joint, command and image streams.

Verify before adoption

Verify clock alignment, dropped frames, action semantics and retention of failed episodes.

03

Learning reference

Publicly available

Project links connect the hardware/data system to ACT and later imitation-learning work.

Verify before adoption

A paper result belongs to its exact dataset, code, hardware and evaluation protocol.

04

Platform evolution

Publicly available

Later research describes ALOHA 2, Mobile ALOHA and ALOHA Unleashed.

Verify before adoption

Do not assume later hardware, models or capabilities are present in the original repository.

Layered licence reading

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

Asset layerKnown statusAdoption decision
Original repository code

Repository identifies MIT.

Retain notices and identify imported ROS and third-party code separately.

Hardware & components

Tutorial links, commercial arms, cameras and documents have separate terms.

A code licence does not grant hardware design, supply, trademark or commercial-use rights.

Demonstration data

Episodes can contain operators, environments and third-party objects.

Define notice, consent, retention, access and redistribution before recording.

Later systems

Publication access does not imply release of every model, dataset or hardware asset.

Use only assets explicitly released under their stated terms.

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

    Configuration graph

    Evidence to preserve

    OS, ROS, workspace commit, arm/controller model, cameras, ports, network and calibration.

    Stop condition

    Stop if tutorial generations or incompatible configurations are mixed.

  2. 02

    Synchronized capture

    Evidence to preserve

    Measured stream rates, timestamps, latency, missing-frame policy and replay comparison.

    Stop condition

    Do not train if observation/action alignment is unknown.

  3. 03

    Operator safety

    Evidence to preserve

    Motion limits, guarded pinch zones, emergency stop, recovery and supervision.

    Stop condition

    Do not place people inside an unvalidated bimanual motion envelope.

  4. 04

    Policy evaluation

    Evidence to preserve

    Held-out tasks, repeated trials, interventions, failure taxonomy and hardware-condition log.

    Stop condition

    A curated success video cannot replace failure-inclusive evaluation.

FUURAA analysis

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

01

Best use

Understanding how bimanual teleoperation, sensing, data collection and imitation learning form one experimental system.

02

Main risk

The ALOHA name spans generations; readers may silently combine original code, later hardware and newer model results.

03

FUURAA judgement

Preserve the operator–sensor–robot–dataset–policy lineage. Every modern substitution creates a new configuration that must earn its own evidence.

Applicability boundary

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

Can support

  • Bimanual teleoperation research
  • Synchronized demonstration collection
  • Imitation-learning architecture analysis

This page cannot replace

  • A current integrated production build package
  • Human-contact robot safety validation
  • Evidence for later variants absent from cited sources
Continue into the robot product development pathFrom requirements and evidence to a bounded prototype →