Skip to content
Saved

Legend

Every word we
apply to a robot.

52 terms make up every classification on this site. Each one is defined here, and each one carries the number of the 582 published records it currently describes.

Autonomy Ladder

Five levels describing how independently a machine operates. Assigned from documented evidence — owner reports, manufacturer documentation, published footage — never from a marketing claim. Every published robot sits on exactly one rung.

LevelWhat it meansRobots on this rung, of 582 published
  1. I

    Manual Automation

    Requires continuous human control for every action. It may automate a physical motion — motorized scrubbing, powered cutting — but cannot navigate, plan, or decide. The human is the operator throughout.

    6 robots 1.0%

  2. II

    Assisted Autonomy

    Handles basic tasks but needs frequent intervention. It may follow simple paths or a schedule, but it gets stuck, misses areas, or needs repositioning. Effort is reduced, not removed.

    176 robots 30.2%

  3. III

    Conditional Autonomy

    Works independently in familiar conditions and needs help with edge cases. It maps, avoids common obstacles, and finishes the job in a known space. New furniture, darkness, or a tangled cord can stop it.

    344 robots 59.1%

  4. IV

    Environmental Autonomy

    Adapts to its environment with minimal input. It reroutes around new obstacles, adjusts to changing conditions, and recovers from most errors unaided. Human involvement is occasional and mostly maintenance.

    56 robots 9.6%

  5. V

    Generalized Autonomy

    Operates fully independently across diverse environments and conditions, requiring no intervention for its core function.

    No consumer robot has met this definition.

The full Autonomy Ladder methodology

Robovations Score

A 0–100 composite describing how much of its job a machine does on its own, and whether it keeps doing it. Nine inputs, all signed the same way, so a higher number always means more capable, more dependable, or less demanding.

What it is made of

Every figure below is a share of the finished 0–100 score, not a share of its pillar — the nine add up to 100.

40%Capability
40%Dependability
20%Ownership burden

What it does without you.

  • Task completion16%
  • Coverage and navigation10%
  • Edge-case recovery10%
  • Operating envelope4%

Whether it keeps doing it.

  • Field failure rate16%
  • Software stability14%
  • Support longevity10%

What it asks of you.

  • Routine maintenance12%
  • Consumables8%

Two things are deliberately missing from that list: price, and what the robot does with your data. We publish both next to the score instead of inside it. A cheap robot earns no points for being cheap. A robot that works without a cloud connection earns none for that either.

What a number means

Five bands across the 0–100 range.

Deficient0–393 robots0.7% of scored
Limited40–5454 robots12.4% of scored
Capable55–69209 robots47.9% of scored
Strong70–84164 robots37.6% of scored
Exemplary85–1006 robots1.4% of scored
corpus median 68

436 robots carry a published score. Range 33–90. A band being small is not a defect — a scarce top mark is the point.

A high score has to be earned in all three parts. An average would let a strong part hide a weak one, so a low part caps the whole score. To reach 85, all three parts must be 70 or above. To reach 70, all three must be 55 or above. Miss a cut-off and the score stops just below it. Two more caps work the same way. A robot heading for abandonment cannot reach the top. And a robot that does very little cannot score well just because there is little to maintain.

The full Robovations Score methodology

Human Readiness

Whether a machine is practically ready for an ordinary household — the gap between what it promises and what a typical home gets. A robot can be highly autonomous and still not be ready, if it takes expert knowledge to set up or keep running.

345
23
2
23

Drawn to scale across 393 classified robots. Segments are floored at a readable minimum, so read the share from the register below.

VerdictRobots carrying it, and their share of the 393 classified
  1. 1

    Ready Now

    On sale past the first-90-day window, assessed, with no documented issues and no announced successor

    345 robots 87.8%
  2. 2

    New to Market

    On sale inside the first-90-day window, so the ownership record has not had time to form. Assessed, with the failure evidence still thin rather than absent.

    23 robots 5.9%
  3. 3

    Wait

    A successor is announced and has not shipped; re-read the decision when it arrives

    2 robots 0.5%
  4. 4

    Known Issues

    Documented problems stand on the record; read the case before buying

    23 robots 5.9%

The full Human Readiness methodology

Tracker vocabulary

Three vocabularies classify every monitoring entry: what kind of event it was, which way it moved, and what it was about.

Tracker entries by type and direction.
each row: entry typeeach column: direction Major advance285 entries Incremental576 entries Lateral233 entries Setback162 entries All types1,256 entries
Announcement 11845% 10941% 3814% · 265
Demo 3754% 2943% 23% · 68
Release 4412% 29276% 4612% · 382
Update 2517% 9768% 107% 118% 143
News 6115% 4912% 13634% 15138% 397

Every published tracker entry carries one type and one direction. Cells are counts; a blank cell is a combination that has never occurred.

Entry type

Announcement265 entries
The manufacturer has publicly stated something — a feature, a timeline, a partnership — that has not been independently verified or shipped.
Demo68 entries
A working prototype or demonstration has been shown, but the capability is not available to buy.
Release382 entries
A product or major feature has shipped and is available to consumers.
Update143 entries
A firmware or software update has changed how the robot behaves. May be positive, neutral, or negative.
News397 entries
Industry news affecting a tracked product that fits no other type — recalls, regulatory action, acquisitions, discontinuation.

Direction

Major advance285 entries
A significant step forward that meaningfully changes the robot’s capabilities, autonomy, or consumer readiness.
Incremental576 entries
A small improvement that refines existing functionality without changing the overall assessment.
Lateral233 entries
Neither clearly positive nor negative. May add features while introducing new issues, or shift capability from one area to another.
Setback162 entries
A change that degrades performance, reliability, or the ownership experience. Includes update regressions, feature removals, and newly discovered issues.

Subject

Every entry also records what the change was about. This vocabulary is applied to all 1,256 entries and is not defined on the live glossary.

Commercial752 entries
A change in availability, price, market, or the company behind the product.
Hardware233 entries
A physical change: components, chassis, sensors, dock, or build.
Software179 entries
A change delivered in firmware or the app.
Autonomy67 entries
A change in how independently the machine operates — the axis the Autonomy Ladder measures.
Safety25 entries
A change bearing on physical, electrical, or data safety, including recalls.

Evidence state

How well corroborated a record is, and where the product sits in its life. This is not a quality judgment — it describes the evidence behind the entry, not the machine.

88
211
117
47
46
73

Drawn to scale across 582 published robots. Counts below are robots.

Verified88 robots
Specs and classification corroborated across 5+ independent sources
Provisional211 robots
Specs on file from 2-4 independent sources; some gaps in evidence
Pending117 robots
Pre-release or fewer than 2 independent sources reviewed
Pre-Release47 robots
Announced but not shipping
Not Consumer-Ready46 robots
Commercial, industrial, or not suitable for consumers
Discontinued73 robots
No longer sold

General terms

Words used throughout the publication that describe what we do rather than what a machine is.

Classification
Categorizing a robot by autonomy level, readiness, and product category. Classification is an objective assessment from documented evidence. It is not a recommendation, an endorsement, or a buying guide.
Evaluation
A structured assessment of one robot against the published methodology. Produces a Robovations Score and documents what the machine does well, what it does not, and on what evidence.
Comparison
A side-by-side reading of two or more robots, structured around an ownership goal rather than a ranking. Comparisons present trade-offs and do not declare a winner.
Hype vs. reality
An assessment separating a manufacturer’s marketing claims from documented performance. Recorded where the two materially diverge.
Unknown
Printed wherever a value has not been established from evidence. It is a real state, not a missing one, and it is never replaced with an estimate.

Definitions change as the frameworks change. The methodology documents are the authority: see how we classify. For reference pages on autonomy vocabulary, category guides, and claim decoders, see the Robot Autonomy Library. Machine-readable site information is at llms.txt.