Industries / Construction and property / Commercial and residential construction / Flooring
Tile and stone
Tile setters bond ceramic, porcelain and stone tile to a floor with cement mortar or adhesive, fill the joints with grout, and leave soft movement joints so the floor can shrink and swell without cracking.
4
Milestones achieved, from c. 825 BCE to 2017.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
c. 825 BCE
A floor of pebbles
At Gordion, the Phrygian capital in central Türkiye, builders paved the hall of Megaron 2 with a pebble mosaic nearly 11 by 10 m. The University of Pennsylvania's Gordion project, which is conserving its 33 surviving pieces, dates it to about 825 BCE and calls it the oldest decorated floor mosaic ever discovered.
Credit
Phrygian builders at Gordion; the Gordion Project
1957
Thin-set mortar without soaking
Herman B. Wagner filed a patent in September 1957, granted in May 1960, for a portland cement mortar with methyl cellulose added to hold its water. Before this, tile went into "heavy, thick, and generally multiple" mortar beds, and porous tile had to be soaked first. The patent's "dry-setting" mortar let dry tile be set on a thin layer, in one example about a quarter of the old thickness and weight, and on gypsum board or plaster. The Tile Council of North America (TCNA) now treats dry-set, thin-set and drybond mortar as the same thing, covered by the standard ANSI A118.1, and polymer-modified versions, still called "latex" mortars, by A118.4, A118.11 and A118.15.
Credit
Herman B. Wagner, Tile Council of America
1987
Uncoupling membrane
Schlüter-Systems introduced DITRA, a membrane laid between the subfloor and the tile so the two can move separately. The company says it copies the old sand layer under tile, which bonding tile straight to the base had removed. TCNA describes the failure this guards against: tile that heaves up, or "tents", is often a sign that too few movement joints were used, and the slab's shrinkage puts the tile under compression.
Lessona tile layer bonded hard to a slab needs a planned way to move, through membranes, movement joints or both.
Credit
Werner Schlüter, Schlüter-Systems, Iserlohn, Germany
2017
Rules for thin porcelain panels
In November 2017 ANSI released two standards: A137.3 sets minimum properties and grading for gauged porcelain tiles and panels, which are porcelain made to a set thin thickness, sometimes with mesh bonded to the back, and A108.19 sets how to install them indoors on a thin bed of modified dry-set mortar. An October 2025 edition added A108.20 for exterior installation.
Credit
ANSI Accredited Standards Committee A108, with TCNA as secretariat
Watch for
Tile-laying robot sold abroad
Partner Robotics, a Chinese start-up, says its PavePal P900 robot lays indoor floor tile "five to six times faster than human labor", that it has finished nearly 100,000 square metres of tiling since mid-2025, and that it has sold robots in Europe, North America, the Middle East and Asia. It also says its controls reduce hollow spots under the tile. No outside test is cited.
Achieved whenA contractor or test lab outside the company publishes measured daily output and installed quality, such as mortar coverage and lippage (the height step between neighbouring tiles), for a real job.
Claimed by
Partner Robotics, November 2025
Robot for heavy tiles
A team at the Technological and Higher Education Institute of Hong Kong (THEi) built a floor-tiling robot for large tiles up to 30 kg. It stands on six electric legs, lifts tiles with four suction cups and finds them with a YOLOv8 vision model. The team reports 30 on-site trials with angle held within ±0.001° and position within 0.5 mm, and lower strain in 10 workers, whose heart rates were 61 to 84 beats per minute with the robot against 90 to 117 by hand.
Achieved whenThe results appear in a peer-reviewed paper, or a contractor outside THEi uses the robot on a job and reports the outcome.
Claimed by
THEi, March 2026
Open
01
Mortar coverage from a photo
A phone tool for a check setters already make by hand. After a freshly set tile is lifted, the tool photographs the back of the tile and the bed it came off, measures the share of the area that touched mortar, and flags bare corners and edges.
The briefCredit · unclaimed02
Tiling robots, tested outside
A public test of tile-setting robots against hand crews on the same floor: one test room, the same tile and mortar, and the same measurements for every run, namely square metres set per hour of crew time, mortar coverage on lifted tiles, lippage at the joints, and every time a person had to step in.
The briefCredit · unclaimed
Open 01 · the brief
Mortar coverage from a photo
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A phone tool for a check setters already make by hand. After a freshly set tile is lifted, the tool photographs the back of the tile and the bed it came off, measures the share of the area that touched mortar, and flags bare corners and edges.
Why it is still open
ANSI A108.5 requires an average mortar coverage of at least 80% (95% for exteriors and showers), judged on at least three lifted tiles, and the person lifting the tile judges coverage by eye. Searches found guidance on doing this by hand but no tool that measures it from images, and tiling-robot makers now claim fewer hollow spots that nobody outside the companies can yet measure.
Built with what exists
A phone camera; a printed scale card; an open image-segmentation model such as Meta's Segment Anything 2; OpenCV and Python.
First step
Set 20 tiles on scrap cement board with three trowel notch sizes, lift them, photograph each, and grid-count them by hand.
Done looks like
On 60 tiles set on test boards with different trowels and then lifted, the tool's coverage figure falls within 5 percentage points of a hand grid count (a clear grid laid over the tile back, counting the squares with mortar) on at least 54, with photos, counts and code published.
Open 02 · the brief
Tiling robots, tested outside
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of tile-setting robots against hand crews on the same floor: one test room, the same tile and mortar, and the same measurements for every run, namely square metres set per hour of crew time, mortar coverage on lifted tiles, lippage at the joints, and every time a person had to step in.
Why it is still open
Every speed figure comes from the robot's makers. Partner Robotics says its PavePal P900 is five to six times faster than human labor, and the THEi robot's results come from its own team; neither names an outside tester.
Built with what exists
Robots lent or rented by their makers; a hand crew paid at the local rate; time-lapse cameras; the mortar-coverage check above or a hand grid count; a straightedge and feeler gauges; a test protocol published before the first run.
First step
Write a two-page test protocol and send it to Partner Robotics and the THEi team, asking whether they will take part.
Done looks like
At least two robots and two hand crews each tile the same 40 m² test floor three times, with every result, video and lifted-tile photo published, including the runs that failed.