An open metaverse for anything you can automate
MetaTwinMaker is an open platform for digital twins of physical things, and the metaverse they live in. A robot vacuum, a self-driving car, a chicken coop, a beehive, the sprinklers on your lawn: if sensors can read it and rules can run it, its twin tracks every variable and shows you every decision. Run a season of your rules on the twin, judge them by what they produce (more eggs, more honey, a greener lawn), then trust them with the real thing.
Twin anything
Any process with something to measure and a decision to make can have a twin: a machine on a line, or a hive in the garden. Some of these you can play with today; the rest are on the drawing board, and we'd love your help building them.
Try it below
Outcome More eggs, every hen home
Daylight, hens on the roost, the temperature outside. The door closes at dusk, but never on a hen that's still out.
Try it below
Outcome More honey, fewer swarms lost
Weigh the hive, listen to the hum, feel the brood warmth. A sudden drop in weight means a swarm, and the twin tells you before the neighbours do.
Try it below
Outcome A greener lawn on less water
Soil moisture, rain forecast, the sun on each zone. Water only what's thirsty, and see why it skipped Tuesday.
Try it below
Outcome Soles that hold, leather that never runs out
A shoemaker's bench: the air, the stock, the order book. Glue when it will hold, stitch when it won't, and order hides before the last one is cut.
Try it below
Outcome Clean floors, nobody stuck under the sofa
The vacuum, the mop, the mower, the window-washer. Their twins learn where the sofa is before the real ones find out the hard way.
Drawing board
Outcome Every chore done, no collisions
A crew of house bots, one job each. Choreograph them in the twin so the mop never chases the vacuum into the same doorway.
Drawing board
Outcome Safe miles
Drive a thousand laps before breakfast. Throw rain, potholes and a runaway football at the twin, not at the car.
Drawing board
Outcome Parcels landed, batteries home
Plan the route, rehearse the landing in a crosswind, and crash a few times for free.
Live now
Outcome Every part in the tray
Six joints and a moving belt. Teach the pick on the twin, then watch every part land in the tray.
Live now
Outcome More good parts, less downtime
Servo axes, presses, mills, lathes, cranes and a whole packaging line, running today in the motion lab. Or try a conveyor sorting parts below.
What's in a twin
A 3D model of the real thing with every part that matters named: the joints that move, the door that shuts, the valve that opens.
Every sensor is a variable you can see: position, light, weight, moisture, temperature. Nothing is decided on a number you can't look at.
Real physics, and the weird stuff too: jams, drift, dead sensors, a hen that won't come in. The twin is only useful if it surprises you the way the real thing does.
Rules you can see
Automation is only fun when you trust it. In the metaverse, the rules are written down where you can read them, and every decision comes with the rule that made it and the numbers it was made on. Try it on a chicken coop, a beehive, the lawn sprinklers, a shoemaker's workshop, a robot vacuum or a conveyor sorting parts: change the conditions, watch the twin decide, and walk around it in 3D.
Lights outWaiting
When daylight is under 20 lux and all 6 hens are on the roost, then close the door.
Now: daylight 120 lux is not under 20.
Wait for stragglersWaiting
When daylight is under 20 lux and a hen is still outside, then keep the door open and send an alert.
Now: daylight 120 lux is not under 20.
Good morningFiring
When daylight is over 50 lux, then open the door.
Now: daylight 120 lux is over 50.
Warm roostWaiting
When it is under 2 °C outside (off again over 5 °C), then switch the heat lamp on.
Now: 8 °C is over 5.
DoorOpen
Heat lampOff
No alerts.
{
"schema": "metatwinmaker.coop.decision/1",
"source": "twin",
"inputs": {
"lux": 120,
"hens": 2,
"tempC": 8
},
"function": "coop.rules@1",
"output": {
"door": "open",
"lamp": false,
"alerts": []
},
"fired": [
"good-morning"
]
}Swarm watchWaiting
When the hum climbs over 350 Hz, then alert the beekeeper to split the hive.
Now: the hum is 240 Hz, not over 350.
Room to growWaiting
When the hive weighs over 45 kg, then ask for another honey box.
Now: the hive weighs 32 kg, not over 45.
Feed the colonyWaiting
When the hive weighs under 20 kg (off again over 24 kg), then open the feeder.
Now: 32 kg is over 24.
Keep the brood coolWaiting
When it is over 36 °C inside (off again under 34 °C), then open the vent and run the fan.
Now: 35 °C is between 34 and 36, so the vent stays shut.
Vent and fanOff
FeederShut
No alerts.
{
"schema": "metatwinmaker.hive.decision/1",
"source": "twin",
"inputs": {
"humHz": 240,
"weightKg": 32,
"insideC": 35
},
"function": "hive.rules@1",
"output": {
"vent": false,
"feeder": false,
"alerts": []
},
"fired": []
}Water when dryWaiting
When the soil is under 25% moisture (until it is over 35%), then run the sprinklers.
Now: the soil at 30% is between 25 and 35, so the lawn stays watered.
Skip for rainWaiting
When over 5 mm of rain is forecast, then hold the water and let the rain do it.
Now: 0 mm of rain is forecast, not over 5.
Wait for the windWaiting
When the wind is over 25 km/h, then hold the water so it doesn't blow onto the path.
Now: the wind is 8 km/h, not over 25.
Soggy soilWaiting
When the soil is over 50% moisture, then alert the gardener to check for a leak.
Now: the soil at 30% is not over 50.
SprinklersOff
LawnWatered
No alerts.
{
"schema": "metatwinmaker.lawn.decision/1",
"source": "twin",
"inputs": {
"soilPct": 30,
"rainMm": 0,
"windKmh": 8
},
"function": "lawn.rules@1",
"output": {
"valve": false,
"thirsty": false,
"alerts": []
},
"fired": []
}Dry air for the glueWaiting
When the workshop is over 65% humidity (off again under 55%), then run the dehumidifier.
Now: the workshop is at 50%, under 55.
Glue only when it will holdWaiting
When the workshop is over 75% humidity, then stitch uppers instead of gluing soles.
Now: the workshop is at 50%, not over 75.
Reorder leatherWaiting
When under 3 m² of leather is left (until over 5 m² is in), then order more hides.
Now: 6 m² is in stock, over 5.
Busy weekWaiting
When more than 12 pairs are waiting, then alert the shoemaker to quote later dates.
Now: 5 pairs are waiting, not more than 12.
DehumidifierOff
At the benchGluing soles
Leather orderNone
No alerts.
{
"schema": "metatwinmaker.shoes.decision/1",
"source": "twin",
"inputs": {
"humidityPct": 50,
"leatherM2": 6,
"pairs": 5
},
"function": "shoes.rules@1",
"output": {
"dehumidifier": false,
"task": "glue",
"ordered": false,
"alerts": []
},
"fired": []
}Clean when dirtyWaiting
When the floor is over 40% dirty (until it is under 15%), then go out and vacuum.
Now: the floor is 25% dirty, between 15 and 40, so the robot stays put.
Charge when lowWaiting
When the battery is under 20% (until it is charged over 80%), then go back to the dock and charge.
Now: the battery is at 90%, over 80.
Empty the binWaiting
When the bin is over 85% full, then stop and ask for it to be emptied.
Now: the bin is 30% full, not over 85.
Deep cleanWaiting
When the floor is over 70% dirty, then turn the suction up to turbo.
Now: the floor is 25% dirty, not over 70.
RobotOn the dock
SuctionOff
No alerts.
{
"schema": "metatwinmaker.vacuum.decision/1",
"source": "twin",
"inputs": {
"batteryPct": 90,
"binPct": 30,
"dirtPct": 25
},
"function": "vacuum.rules@1",
"output": {
"activity": "docked",
"turbo": false,
"alerts": []
},
"fired": []
}Keep paceWaiting
When over 40 parts a minute are coming in (until under 25), then run the belt fast.
Now: 20 parts a minute are coming in, under 25.
Sort trueWaiting
When over 8% of parts are bad, then keep the belt at normal speed, so the pusher hits every bad part.
Now: 3% of parts are bad, not over 8.
Hold a bad batchWaiting
When over 15% of parts are bad (until under 8%), then stop the line and call quality.
Now: 3% of parts are bad, under 8.
Spare reject binWaiting
When the reject bin is over 85% full, then send rejects to the spare bin and ask for it to be emptied.
Now: the reject bin is 30% full, not over 85.
BeltNormal
Rejects toReject bin
No alerts.
{
"schema": "metatwinmaker.sorter.decision/1",
"source": "twin",
"inputs": {
"partsPerMin": 20,
"badPct": 3,
"rejectBinPct": 30
},
"function": "sorter.rules@1",
"output": {
"belt": "normal",
"rejectsTo": "reject",
"alerts": []
},
"fired": []
}Outcomes
Closing a door at dusk is an action; a full egg basket and every hen home safe is the outcome. A twin runs a whole season of your rules before you live it, on the same weather as doing it the old way, so you can judge the rules by what they produce.
A timer and a calendar The twin's rules
Season 1
+48 eggs collected (+21%)
The door waits for stragglers and shuts at dusk, the lamp keeps the roost off the cold, so the hens keep laying and the fox finds nobody outside.
Eggs collected
225 273Hens home at the end
5 of 6 6 of 6Nights shut outside fewer is better
10 0Cold nights on the roost fewer is better
29 0Heat lamp hours the price of the rules
0 h 924 h+18.3 kg honey harvested (+68%)
The hive's weight and hum flag swarm preparations in time to split it, its weight says when to add the next box, and low stores say when to feed.
Honey harvested
26.8 kg 45.1 kgSwarms lost fewer is better
2 0Colony strength at the end
90% 99%−9,800 L water used (−22%)
The soil probe says when the lawn is thirsty, the forecast says when rain will do it instead, and the wind says when to wait, so it stays green on less water.
Water used less is better
44,800 L 35,000 LDays the lawn was green
94 of 112 109 of 112Days parched fewer is better
13 2Days waterlogged fewer is better
5 1Water blown onto the path less is better
1,200 L 0 L+7 pairs delivered on time (+13%)
Soles are glued only when the air will let them hold, leather is ordered before the last hide is cut, and a full order book quotes honest dates, so pairs go out right and on time.
Pairs delivered on time
52 59Soles that lifted and came back fewer is better
5 1Pairs delivered late fewer is better
3 0Days without leather fewer is better
1 0Dehumidifier hours the price of the rules
0 h 328 h+4 evenings with a clean floor (+27%)
It goes out when the floor is dirty rather than when the clock says, docks before it runs flat, and asks for its bin to be emptied instead of sweeping with it full.
Evenings with a clean floor
15 of 28 19 of 28Hours with a dirty floor fewer is better
279 h 70 hTimes stranded with a flat battery fewer is better
5 0Hours swept with a full bin fewer is better
27 h 0 hTimes asked to empty the bin the price of the rules
0 6−88 bad parts shipped (−43%)
The line holds itself the minute the camera sees a bad batch, slows down when bad parts are about so the pusher catches them, and switches bins before a full one jams it.
Bad parts shipped fewer is better
207 119Good parts sorted
62,246 63,169Bad parts made fewer is better
4,133 3,210Minutes jammed by a full reject bin fewer is better
179 min 0 minMinutes held for quality the price of the rules
0 min 181 minAn illustrative model with plausible numbers, not a measured result: six hens, one hive, a 100 m² (1,100 sq ft) lawn, a one-person workshop, one robot vacuum and one sorting line, the same weather and the same luck for both runs, so the rules are the only difference. A twin's job is to run this on your own numbers.
Same shape, both worlds
A twin is an emulation: it runs the same function on the same inputs as the real controller, and writes its answer as the same JSON. Lay the two records side by side and the real world can be checked against the twin, field by field.
The twin and the real coop get the same inputs and run the same rules.
Their outputs are compared field by field. A mismatch is a finding: a stuck door, a drifting sensor, a rule that changed.
Every record has the same shape, so a classifier trained on the twin's records reads the real ones without translation, and quickly puts a probability on each outcome.
source: "twin"{
"schema": "metatwinmaker.coop.decision/1",
"source": "twin",
"inputs": {
"lux": 12,
"hens": 6,
"tempC": 1
},
"function": "coop.rules@1",
"output": {
"door": "closed",
"lamp": true,
"alerts": []
},
"fired": [
"lights-out",
"warm-roost"
]
}source: "coop-01"{
"schema": "metatwinmaker.coop.decision/1",
"source": "coop-01",
"inputs": {
"lux": 12,
"hens": 6,
"tempC": 1
},
"function": "coop.rules@1",
"output": {
"door": "open",
"lamp": true,
"alerts": []
},
"fired": [
"lights-out",
"warm-roost"
]
}5 of 6 fields match
inputs.luxinputs.hensinputs.tempCfunctionoutput.doortwin closed, coop openoutput.lampSame inputs, same function, different output: the twin closed the door and the coop's didn't. That is a finding, not a mystery. Check the door motor.
From the mismatched record, a classifier ranks the causes by probability, so you check the likeliest first. The same goes for outcomes still to come: a swarm this week, a hen that won't come in, a bearing on its way out.
Example output, to show the shape of the answer.
How it works
That's the whole game, and the twin is what makes the risky parts cheap. Nothing catches fire, no hen gets locked out, and the undo button always works.
Model the thing and name the parts that move, open, shut or switch.
Give it the variables its real sensors read: light, weight, moisture, position.
Write the rules in plain terms, or program it the way you'd program the real one.
A cold snap, a stuck valve, a hen who stays out late. Go on, we'll wait.
Judge the rules by what they produce, and check the real thing against the twin.
An open platform
Twins are useful for far more than one thing. Here's what people can do with them, and what we're building the platform for.
Labs on machines you could never fit in a classroom, with progress and quizzes.
Try your code on the twin before it ever meets the hardware.
Challenges, races and puzzles where the physics is real and the stakes aren't.
Lay out a robot fleet or a factory floor and watch it run before you buy a thing.
Mirror the real coop, hive or line: every reading on its twin, every decision in its log.
Reproduce a fault on the twin of the thing that has it, then try the fix.
Every world publishes what it holds at /twins.json: each twin's 3D model
and joints, its camera views, its troubleshooting scenarios and its tags, described
by a JSON schema and readable from anywhere.
Tags come from a shared vocabulary, so twins from different worlds can be found and grouped together. This page is built from those catalogs, and your app can be too.
GET https://motion.metatwinmaker.com/twins.json
{
"site": { "name": "MotionMeta", "parent": "https://metatwinmaker.com" },
"twins": [
{
"id": "robot",
"name": "Robot cell",
"model": { "url": ".../models/robot.glb", "joints": […] },
"scenarios": […],
"classification": { "category": "robotics", … }
},
…
]
}Worlds
A world is a corner of the metaverse on its own subdomain, with the physics and the twins its field needs. metatwinmaker.com ties them all together. The first one is open.
Live · Motion control
Logix integrated motion: servo axes, coordinated motion, camming and gearing, robot and CNC control, taught on machines you can run, break and fix in the browser.
Lab hall. An 18 by 12 m lab hall with every twin at its own station; visitors walk it in first person and step up to a station to use it.
Your world next?
Robot butlers, a greenhouse, an apiary, a self-driving go-kart, the model railway in
the attic. Anything with something to measure and a rule to follow can have a
world. Publish a
twins.json and its twins show up on this page.
Play now
Straight from the motion lab and running live in your browser. Pick one, drive it, then break it on purpose.
From the catalog at motion.metatwinmaker.com.
One servo axis driving a ballscrew carriage: tune it, move it, fault it and reset it.
Axes: Axis
A tilting ladle pouring copper into two indexing casting wheels, with gearbox faults to find.
Axes: Ladle tilt, Wheel A, Wheel B · 10 scenarios
An X–Y gantry running coordinated lines and arcs, with its path plotted as it goes.
Axes: X, Y
Eight movers on a loop of linear motor segments, with a switch into a machine lane and an entry merge.
Axes: Carts (eight), Switch · 7 scenarios
A six-axis arm picking parts off a moving belt into a tray, programmed as Logix coordinated motion.
Axes: J1–J6, Belt · 6 scenarios
A mobile robot carrying trays across the hall from the robot cell to the cart track, under a fleet manager.
Axes: Left and right wheels · 5 scenarios
A miniload crane storing and retrieving totes in a two-rack aisle, with mast sway to settle.
Axes: Travel, Hoist, Forks · 5 scenarios
An electric press fitting pins into housings on a six-nest index table, judged by its force–depth curve.
Axes: Press rod, Index table · 5 scenarios
A four-axis wet winder laying carbon fibre on a pressure vessel's liner, its axes cammed to the spindle.
Axes: Spindle, Carriage, Cross-feed, Payout eye, Dancer feed · 5 scenarios
An enclosed three-axis vertical mill running a part program as coordinated motion, cutting a block you can watch.
Axes: X, Y, Z, Spindle, Carousel · 5 scenarios
A two-axis CNC turning centre turning a threaded pin from bar and measuring it against its drawing.
Axes: Spindle, Z, X, Turret, Tailstock quill · 5 scenarios
A pick-and-place delta robot picking products off a moving belt by camera and packing them into trays.
Axes: J1–J3, J6, Conveyor · 6 scenarios
The lab's machines joined into one autonomous line, from the press that makes each product to the warehouse rack that stores it, in a room of its own.
Axes: Press, Cart track, Delta robot, Case packer, Mobile robot, Warehouse crane · 5 scenarios · In the packaging line room
The motion lab is open: robot arms, cranes, mills and a whole packaging line, all waiting for someone to push the wrong button.