BRENNEN

Industries / Food, farming and forestry / Agriculture and farming / Dairy

Milking

Getting milk out of a cow, buffalo, goat, sheep or camel, by hand, by machine or by robot, without harming the udder.

5

Milestones achieved, from 1895 to 2012.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1895

    A pulsator rests the teat

    The first milking machines used steady suction, and attempts from 1851 drew blood and fluid into the teat, by the account in Gale's How Products Are Made. Alexander Shiels of Glasgow patented a pulsator, a valve that switches the suction on and off in a rhythm so the teat gets a rest between pulls; he filed the US patent on 27 September 1895 and it was granted on 21 April 1896. Every milking machine still has a pulsator.

    LessonPulling harder did not work; machines replaced hand milking only after the rest phase and a two-chambered teat cup (1892) were added, by Gale's account.

    patents.google.comencyclopedia.com

    Credit

    Alexander Shiels

  2. 1930

    Cows ride a turntable

    On 13 November 1930 the Borden Company's dairy farm at Plainsboro, New Jersey, started the Rotolactor, a rotating platform that washed and milked 50 cows in 12.5 minutes (Progressive Dairy's account). It was shown at the 1939 New York World's Fair. Rotaries sized for ordinary farms came later: in the 1968–69 season Taranaki farmer Merv Hicks built a walk-on, back-off rotary shed in New Zealand, the start of the rotaries large herds use today.

    agproud.comexhibits.princetonlibrary.orgsciencelearn.org.nznumber8network.co.nz

    Credit

    Henry W. Jeffers, president of Walker-Gordon, who had the idea (Princeton Public Library); Merv Hicks

  3. 1952

    Milking from a pit

    Waikato farmer Ron Sharp built a shed at Gordonton, New Zealand, where cows stand at an angle on raised platforms on each side of a pit, so the milker puts on cups at eye level without bending. This layout, the herringbone, spread fast: by 1965, 69% of new New Zealand sheds and 57% of rebuilt ones used it, and by the early 1980s nearly 68% of the milk collected by the UK's Milk Marketing Board came from herringbone sheds, by the count of Sharp's son Alan.

    sciencelearn.org.nznumber8network.co.nz

    Credit

    Ron Sharp, made an Officer of the New Zealand Order of Merit for services to dairying in 2000

  4. 1992

    Robots milk on demand

    The first commercial milking robots went onto Dutch farms in 1992. A robot box finds the teats and puts the cups on by itself, and each cow walks in to be milked when she chooses. Kees de Koning of Wageningen University writes that prototypes took a decade to become reliable, uptake stayed slow until 1998, and by the end of 2003 about 2,200 farms used robots, more than 85% of them in north-western Europe.

    LessonA European study of 44 farms (Hillerton and colleagues, 2004) found no major problems in switching but no substantial gain in any aspect of cow health, and bacteria counts in the milk often rose somewhat, because cleaning and cooling had to change for milking around the clock.

    icar.org

    Credit

    Dutch makers and researchers; history by Kees de Koning

  5. 2012

    Robots on a rotary

    On the Dornauf family's Gala farm near Deloraine, Tasmania, a 24-bail DeLaval automatic milking rotary started work in February 2012; the trade journal Dairy Exporter reported it as the first on a commercial farm. Robots stand inside the turning platform and each does one job: two clean teats, two put on cups and one sprays teats after milking. About 8–12% of cows went round a second time because a cup had not attached or a quarter of the udder had not milked out.

    futuredairy.com.au

    Credit

    DeLaval with the FutureDairy project at Camden, New South Wales, led by Kendra Kerrisk; the Dornauf family

Watch for

  1. Robots milking in batches

    DeLaval says robot boxes can milk a whole group at fixed times, as a parlor does. It reports that a farm at Cieszymowo, Poland, run by Fortune Sp. z o.o., milks up to 1,500 cows a day with 25 VMS V300 robots, that more than 25 farms used this approach in 2025, and that the number is set to double in 2026. GEA announced its own batch-milking layout, the DairyRobot R9600, in October 2026.

    Achieved whena university or extension study reports labor hours, milk per cow and udder health on batch-milking robot farms against similar parlor farms.

    tetralaval.comgea.com

    Claimed by

    DeLaval (Tetra Laval's report for 2025), 2026; GEA, October 2026

  2. One control room, eighteen robots

    Lely says its Astronaut Max links up to 18 milking robots to one central room that supplies vacuum, cleaning and milk transport, for farms of 500–1,100 cows. It says letting robots keep milking while others are cleaned "significantly increases uptime and milk capacity", but gives no figures. The system was in validation, with a phased rollout from 2026.

    Achieved whenan independent party publishes uptime and milk per robot from working farms.

    lely.com

    Claimed by

    Lely, June 2025

Open

  1. 01

    Open cup-attaching test rig

    A bench test anyone can copy, for the hardest step in robot milking: putting four cups on four teats. A silicone udder on a frame sets the teats at random positions and angles. A desktop robot arm finds them with a depth camera and attaches teat cups under vacuum, with water in place of milk. Release the designs, code, recordings and a scoring sheet, so research and commercial systems can be compared on the same rig.

    The brief
    Credit · unclaimed
  2. 02

    Open milking-machine tester

    A pocket logger that plugs into a milking unit's pulse tube and claw, records the vacuum at least 100 times a second, and works out the pulsation rate (cycles per minute), the pulsation ratio (the share of each cycle spent milking) and the length of each phase, shown on a phone. Publish the hardware, the firmware and a side-by-side test against a commercial tester, so a farmer or local technician can check a machine more often.

    The brief
    Credit · unclaimed

Open 01 · the brief

Open cup-attaching test rig

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A bench test anyone can copy, for the hardest step in robot milking: putting four cups on four teats. A silicone udder on a frame sets the teats at random positions and angles. A desktop robot arm finds them with a depth camera and attaches teat cups under vacuum, with water in place of milk. Release the designs, code, recordings and a scoring sheet, so research and commercial systems can be compared on the same rig.

Why it is still open

Commercial robots are closed, and published prototypes do not release their designs. The nearest work found: Wang and colleagues (Agriculture, 2023) report a 98% attach rate and 21 seconds per cow for a purpose-built six-axis arm; Borla and colleagues (arXiv, 2021) estimated the positions of synthetic teats but did not attach cups; and a 2026 GitHub project on teat detection and disinfection has no published results. No shared test rig, and no open design with a published attach rate, was found.

api.crossref.orgarxiv.orggithub.com

Built with what exists

A desktop six-axis arm such as the UFactory xArm 6, programmed through its open Python SDK; a RealSense D405 depth camera; Ultralytics YOLO to find teats in images; teats cast in platinum-cure silicone from 3D-printed molds; standard teat cups, liners and a portable bucket milking unit for vacuum. A learned policy can be trained from recorded attempts with Hugging Face's LeRobot library (see the Note "No download").

  1. First step

    Cast four silicone teats, mount them on a board with movable positions, and record and label 100 depth images of the teat tips.

  2. Done looks like

    200 filmed attempts across at least 20 teat layouts, including wide, close and tilted teats, with the attach rate per cup, the time per udder and a published tally; and a second team that builds the rig from the files and gets an attach rate within 5 percentage points of the first.

Open 02 · the brief

Open milking-machine tester

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A pocket logger that plugs into a milking unit's pulse tube and claw, records the vacuum at least 100 times a second, and works out the pulsation rate (cycles per minute), the pulsation ratio (the share of each cycle spent milking) and the length of each phase, shown on a phone. Publish the hardware, the firmware and a side-by-side test against a commercial tester, so a farmer or local technician can check a machine more often.

Why it is still open

Testers on sale, such as the VacuScope easy, are closed instruments. The ISO 6690 standard asks for at least 100 samples a second when testing pulsators, and the nearest published low-cost build, by Roca and colleagues (Journal of Agricultural Informatics, 2017), was a computer-driven laboratory rig. Searches of GitHub, OpenAlex and the web found no open design with a published comparison against a certified tester.

real.mtak.huelmicro.com

Built with what exists

A fast absolute-pressure sensor (Roca and colleagues list sensors that respond in under 1 millisecond); an ESP32 microcontroller with Bluetooth; a phone app built with an open framework such as Flutter; a commercial pulsator tester borrowed for comparison.

  1. First step

    Wire one sensor to an ESP32, log a running pulsator at 200 samples a second, and compare one minute of readings with a commercial tester.

  2. Done looks like

    On at least 30 milking units on five or more farms, including bucket milkers, the logger's pulsation rate, ratio and phase lengths agree with a commercial tester within limits written down before testing, with every paired reading published.