BRENNEN

Industries / Consumer goods / Textiles and apparel / Footwear

Assembly

Assembly is the work of making a shoe from a finished upper and a sole: the upper is pulled tight over a last, the foot-shaped form a shoe is built on, and fastened there, which is called lasting, and the sole is then joined to it by stitching, by gluing (called cementing), or by moulding the sole straight onto it.

7

Milestones achieved, from 1858 to 2025.

3

To watch for: announced, not yet confirmed outside.

3

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1858

    Soles sewn on by machine

    Lyman R. Blake of South Abington, Massachusetts, patented a sewing machine on 6 July 1858 (US 20,775) in which the part the shoe rests on sits on an arm that goes inside the shoe, so a needle and looper could do "the sewing of the sole to the upper and insole". In 1859 he sold the rights to Gordon McKay, who made the arm, or horn, movable and sold the result as the McKay stitching machine. Charles Goodyear Jr. of New Rochelle, New York, patented machinery for "Stitching Soles to the Welts" of boots and shoes on 24 January 1871 (US 111,197); a welt is a strip sewn around the bottom edge of the upper, to which the outsole is then stitched. Fiddlebase, a collectors' archive of shoe-machine patents, says the welt machine itself came from two mechanics Goodyear hired, Auguste Destouy and Daniel Mills, and that the welting patent that took Goodyear's name dates from 1875.

    patents.google.comfiddlebase.compatents.google.comfiddlebase.com

    Credit

    Lyman R. Blake and Gordon McKay; Charles Goodyear Jr., Auguste Destouy and Daniel Mills

  2. 1883

    Lasting by machine

    Jan Ernst Matzeliger of Lynn, Massachusetts, received US patent 274,207 on 20 March 1883 for a lasting machine, whose stated object was "to perform by machinery" work then done by hand; he assigned two-thirds of the patent to Melville S. Nichols and Charles H. Delnow. By Lemelson-MIT's account, the machine held the shoe on a last, pulled the leather down around the heel, drove the nails and released the shoe. Lemelson-MIT gives its output as 150 to 700 pairs a day against a skilled hand laster's 50; the National Inventors Hall of Fame gives 700. The Consolidated Lasting Machine Company was formed to build the machine, and United Shoe Machinery later obtained the patent.

    patents.google.comlemelson.mit.eduinvent.org

    Credit

    Jan Ernst Matzeliger

  3. 1953

    Leasing ruled a monopoly

    On 7 February 1899 seven shoe-machinery companies were merged into the United Shoe Machinery Company, and in May 1918 the Supreme Court let the merger stand. In February 1953 Judge Wyzanski of the federal court in Massachusetts found that United supplied more than 75% of the demand for shoe machinery, that it leased its more important machines rather than selling them, on terms that included a 10-year rental and a "full capacity" clause, and that through these leases it reached about 90% of all shoe factories. He held this to be monopolisation under Section 2 of the Sherman Act and ordered United to offer for sale every type of machine it leased and to limit leases to five years with no full-capacity clause, but refused to split it into three companies. The Supreme Court affirmed on 17 May 1954.

    LessonIn 1968 the Supreme Court sent the case back, holding that if ten years of the decree had not achieved its main aims, "the time has come to prescribe other, and if necessary more definitive, means".

    law.cornell.educase-law.vlex.comcase-law.vlex.comlaw.cornell.edu

    Credit

    the US District Court for Massachusetts, in a suit brought by the US government

  4. 1959

    Soles moulded onto uppers

    In direct soling, also called direct injection, the lasted upper is closed into a mould and the sole material is injected against it, so the sole is formed and joined to the upper in one step. DESMA, a machine maker founded in Achim, Germany, in 1946, dates its work on the process with vulcanising presses and moulds to 1953 and its first direct soling machine to 1959, for pasty PVC, with a machine for granules a little later. ECCO writes that direct injection "had originally been developed for the mass production of low-cost shoes" before ECCO used it for its Soft shoe in the early 1980s. In 2020 SATRA, a footwear testing and research organisation, reported DESMA's claim that its systems are involved in making one-tenth of all shoes made worldwide.

    desma.deenter.ecco.comsatra.com

    Credit

    DESMA

  5. 1998

    Nike's factory pledges

    On 12 May 1998 Nike pledged to raise the minimum age at the factories that make its sports shoes to 18, and at its clothing factories to 16. As the Seattle Times listed the pledges of Nike's chairman Phil Knight, every Nike shoe factory would also meet the US Occupational Safety and Health Administration's indoor air-quality standards, and Nike would include non-governmental organisations in factory monitoring.

    LessonA 2001 report by the campaign group Global Exchange found that the projects Knight announced had been "of little benefit to Nike workers", though it said Nike had improved health and safety.

    time.comarchive.seattletimes.com

    Credit

    Nike

  6. 2019

    Robot factories closed

    adidas set up a pilot Speedfactory in Ansbach, Germany, in December 2015 to make shoes by automated production close to buyers, and later a second in Atlanta, both operated by Oechsler. On 11 November 2019 adidas said production at both would end "by April 2020 at the latest" and that the technology would move to two of its suppliers in Asia. CNN reported that the two plants were targeting 1 million pairs a year between them; adidas told CNN it makes more than 400 million pairs a year. Nike's project with the electronics maker Flex, begun in 2015 to partly automate sneaker making in Guadalajara, Mexico, was wound up by early 2019; the Wall Street Journal reported that the robots "struggled to handle the soft, squishy and stretchy parts" of shoes, that gluing soles to uppers proved hard to automate, and that factory staff grew to 5,000, about twice the plan.

    LessonThe Journal named the variety of shoes Nike makes as a central problem; at one point it took the Flex team eight months to automate putting the swoosh on a shoe.

    res.cloudinary.comres.cloudinary.comkvia.comworldfootwear.comtovima.com

    Credit

    adidas and Oechsler; Nike and Flex

  7. 2025

    Asia makes nine in ten

    APICCAPS, the Portuguese footwear association, says in its World Footwear Yearbook 2026 that Asia made 88.7% of the world's 24.6 billion pairs in 2025. The work also moves within Asia: China's share of world exports fell to 61.2% from 69.1% ten years earlier, and Vietnam's rose to 11.2%. Contract makers, the factories that make shoes for brands, work at a large scale: Yue Yuen, majority-owned by Pou Chen of Taiwan, shipped 252.2 million pairs in 2025 for brands including Nike, adidas, Asics, New Balance and Salomon, with about 276,800 people in its manufacturing business, and made 54% of those pairs in Indonesia, 32% in Vietnam and 9% in mainland China, by its own annual report.

    worldfootwear.comwww1.hkexnews.hk

    Credit

    APICCAPS (figures); Yue Yuen and Pou Chen

Watch for

  1. Robots that learn gluing

    Yingshen Intelligence, founded in Hangzhou in 2024, says its ShadowGlue robot learns how to glue shoe parts from videos of skilled workers and adapts to how the material deforms, to the edge path and to the state of the adhesive, and that its ShadowPress robot presses soles and changes its gripper angle and path when a part's position or the shoe's shape varies. In July 2026 it signed agreements with footwear makers in Hai Phong, Vietnam, among them Power Loong (Giày Bảo Long) and Thuận Phi, for "hundreds" of robots worth "tens of millions of yuan". It has not disclosed a delivery timetable or how many robots are installed.

    Achieved whenone of the Vietnamese factories, or a visitor with no stake in Yingshen, reports the robots running on production lines, with sole-bond test results and output compared with hand gluing on the same shoes.

    vietnamnews.vnequalocean.com

    Claimed by

    Yingshen Intelligence, July 2026, in a Media OutReach release reported by Vietnam News and EqualOcean

  2. Four robots prepare uppers

    Abílio P. Carneiro & Filhos, a Portuguese maker of technical outdoor and waterproof footwear, says it built a line of four robots, a dryer and a conveyor in the FAIST research programme that roughs the sides and underside of uppers, meaning it scuffs the surface so glue will hold, and applies primer and adhesive before the outsole is glued on. The company says the line has "significantly improved precision, quality and process consistency".

    Achieved whenthe line runs in regular production and someone outside the company publishes its output and the sole-bond results of its shoes beside those of the hand line.

    worldfootwear.com

    Claimed by

    Alexandre Carneiro, chief executive, in World Footwear, September 2026

  3. Cameras and cobots on outsoles

    Genon, a South Korean AI company, says its camera inspection system passed a production validation for Nike footwear, finding defects in Nike outsoles with accuracy of 99.7% and 99.2% on two styles, and that it has won a project applying "physical AI" to outsole pressing, which combines cobots (robot arms built to work beside people) with cameras. The article does not say who ran the validation or which factory is involved.

    Achieved whenNike or the factory confirms the system is in use, and someone outside Genon reports detection and pressing results on production shoes.

    koreatimes.co.kr

    Claimed by

    Genon, in the Korea Times, October 2026

Open

  1. 01

    Robot pulls upper onto last

    A bench cell that takes a sewn sports-shoe upper, pulls it toe first onto a plastic last held in a clamp, and works the heel down until it is seated, the step before the sole is attached. Two low-cost arms grip and guide the upper; the pull itself comes from a motor-driven linear stage with a load cell, so the force is measured and capped.

    The brief
    Credit · unclaimed
  2. 02

    A public bite-line set

    A free set of 3D scans of at least 200 lasted uppers, each with its bite line, the line on the upper where the top edge of the sole will sit and which marks the limit for roughing and gluing, drawn by a skilled worker and traced in 3D, together with a scoring script that reports how far, in millimetres, a predicted line lies from the drawn one, and a baseline model.

    The brief
    Credit · unclaimed
  3. 03

    Sole bonds from shops

    A public table of sole-bond strength for at least 60 pairs of cemented and direct-injected shoes bought in shops in at least three countries and three price bands, each tested by the peel method in SATRA TM411, with the peel force, where on the shoe it was taken, and a photo of how the bond failed.

    The brief
    Credit · unclaimed

Open 01 · the brief

Robot pulls upper onto last

Credit · unclaimed

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

What to build

A bench cell that takes a sewn sports-shoe upper, pulls it toe first onto a plastic last held in a clamp, and works the heel down until it is seated, the step before the sole is attached. Two low-cost arms grip and guide the upper; the pull itself comes from a motor-driven linear stage with a load cell, so the force is measured and capped.

Why it is still open

KUKA was granted a US patent on 23 July 2024 for a robot that pulls an upper onto a last under force control, with a maximum pull of 100 to 250 N, but has published no trial results. The Wall Street Journal reported in 2025 that Nike's automation project in Mexico was wound up after robots struggled with the soft, stretchy parts of shoes. Searches for robot last insertion, robotic upper-on-last trials and reports on hand work in sneaker factories found the patent and a 2003 research cell at the Jožef Stefan Institute that used a robot to handle lasts beside a standard lasting machine, but no published trial of a robot pulling uppers onto lasts.

patents.google.comtovima.com

Built with what exists

Two low-cost arms and Hugging Face's LeRobot library to record steered demonstrations and train on them (see the Note "No download"); for the trained policy, NVIDIA's GR00T 2, which learned tasks from five demonstrations in NVIDIA's own test (see the Note "Five demonstrations per task"), or Runway's open Praxis-1 weights (see the Note "Open robot model weights"); a stepper-driven linear stage with a load cell, because low-cost arms are not built for pulls of the size in KUKA's patent; plastic lasts and sewn uppers from a shoemaking school or sample room; a camera.

  1. First step

    Fit a load cell to a strap at the heel and record the force and time a person uses to pull ten uppers onto one last, filming each pull.

  2. Done looks like

    20 filmed attempts on one upper style and size, with a published tally of uppers seated with no person stepping in (heel down on the last and the back seam within 3 mm of a line drawn on the last), and the peak pull force and time for each attempt beside a person's results on the same 20 uppers.

Open 02 · the brief

A public bite-line set

Credit · unclaimed

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

What to build

A free set of 3D scans of at least 200 lasted uppers, each with its bite line, the line on the upper where the top edge of the sole will sit and which marks the limit for roughing and gluing, drawn by a skilled worker and traced in 3D, together with a scoring script that reports how far, in millimetres, a predicted line lies from the drawn one, and a baseline model.

Why it is still open

Robotic gluing needs this line, and in factories a person still draws it: a Nike patent filed in 2019 and granted in November 2022 describes how "a worker manually traces" it on the lasted upper. Research groups test their systems on their own equipment, such as Min Gyu Kim and colleagues at the University of Seoul in 2022, who used "an automated production testbed". Searches for public bite-line datasets and for lasted-upper point clouds, including on Hugging Face and Kaggle, found only patents and papers.

patents.google.compure.uos.ac.kr

Built with what exists

A handheld structured-light 3D scanner; lasted uppers, not yet soled, from a factory or shoemaking school; a silver marking pen; Open3D to process the scans; an open point-cloud model such as PointNet++ as the baseline; Python.

  1. First step

    Borrow ten lasted uppers from a shoemaking school or small factory, scan them, have a worker draw the bite line, and scan them again.

  2. Done looks like

    200 uppers from at least five styles and three sizes, released under CC BY 4.0, with two workers marking the same 50 uppers to measure how far people disagree, and the baseline's mean and 95th-percentile error reported against that.

Open 03 · the brief

Sole bonds from shops

Credit · unclaimed

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

What to build

A public table of sole-bond strength for at least 60 pairs of cemented and direct-injected shoes bought in shops in at least three countries and three price bands, each tested by the peel method in SATRA TM411, with the peel force, where on the shoe it was taken, and a photo of how the bond failed.

Why it is still open

SGS, a testing company, wrote in 2017 that "on average, over 15% of footwear fails in testing because of poor sole bonding". Those tests are run for brands and factories, and their results are not published shoe by shoe. Searches for sole-bond surveys by consumer testers such as Which?, Stiftung Warentest and Consumer Reports found none.

ca.marketscreener.comsatra.com

Built with what exists

A tensile testing machine of the kind many university materials labs own; a copy of SATRA TM411; a knife or band saw to cut test strips; a spreadsheet.

  1. First step

    Buy ten pairs of low-priced cemented trainers, cut strips as TM411 sets out, and test them on a university tensile tester.

  2. Done looks like

    60 pairs, three strips from each shoe (toe, waist and heel), each with a peel force in newtons per millimetre of width and a failure type, and all load curves and photos published, grouped by construction, price band and country of manufacture.