Industries / Construction and property / Heavy civil / Paving
Compaction
Compaction presses soil, stone and fresh asphalt into a denser, stronger layer, mostly with heavy rollers.
5
Milestones achieved, from 1867 to 2020.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1867
Steam rolls the road
A road roller is a heavy machine on wide, smooth steel wheels, called drums, that presses each new layer down. Thomas Aveling of Rochester, Kent, whose firm became Aveling and Porter when Richard Porter joined it in 1862, tested a steam roller in 1865 in Hyde Park, London, on Military Road in Chatham and at Star Hill in Rochester, by the account of Grace's Guide to British Industrial History. In 1867 the firm built a 30-ton steam roller for the Liverpool Corporation, and the Science Museum Group holds a photograph from the journal Engineering of October 4, 1867, captioned as the firm's first steam road roller. Its rollers went to the US Arsenal in Philadelphia in 1868 and to Central Park in New York in 1869.
Credit
Thomas Aveling and Richard Porter, Aveling and Porter
1956
Drums that shake
A vibratory roller turns an off-center weight inside its drum to make the drum vibrate, and it compacts deeper layers than weight alone. A column for the Historical Construction Equipment Association says the first vibratory drum compactors were single drums towed behind crawler or wheel tractors, that who made the first and when is unknown, and that the maker Essick built them as early as 1956. Cedarapids was offering a towed vibratory compactor on rubber tires by 1953. Self-propelled single-drum vibratory rollers arrived in force in the mid-1960s, and tandem rollers, with two drums on one machine, soon followed.
Credit
no single inventor is known; early makers include Essick and Cedarapids
1980
The roller measures the ground
Heinz Thurner and Åke Sandström of Geodynamik, a Stockholm firm, found in 1979 that the way a vibrating drum moves depends on the stiffness of the soil beneath it. In 1980, at the International Conference on Compaction in Paris, they presented "a new device for instant compaction control." Geotechnical researchers at TU Wien call it the first system to monitor compaction from the roller itself while the work goes on, and the idea behind the intelligent compaction systems that followed. Their US patent senses the drum's movement and judges the degree of compaction from the timing or the relative size of its swings.
Credit
Heinz Thurner and Åke Sandström, Geodynamik H. Thurner AB
2011
Rollers that map the job
Intelligent compaction (IC) puts an accelerometer, a motion sensor, on the roller drum, satellite positioning on the machine, and a color map in front of the operator that shows where the roller has been and how stiff the ground responded. In 2011 the US Federal Highway Administration (FHWA) completed a three-year pooled-fund study with state transportation departments, with 16 IC demonstrations in 13 states, and followed it with generic IC specifications for soil and asphalt and a data program called Veda. The Minnesota Department of Transportation said it aimed to cut its acceptance testing by two-thirds with IC.
Lessonon asphalt, IC readings matched core densities inconsistently, and on one test bed ran the wrong way, because the drum senses the stiffness of the whole pavement and the ground beneath while density is measured in the top layer only; the study found IC most useful for finding weak spots before paving and for making sure every part of a layer was rolled.
Credit
FHWA (Lee Gallivan), The Transtec Group (George Chang) and the participating states
2020
Rollers with nobody aboard
Chinese machine makers have put fleets of driverless rollers to work on expressway jobs, steered by satellite positioning and watched from a control room. XCMG says its autonomous roller completed tests at the end of May 2020 on the Panzhihua–Dali (Panda) Expressway, working to within 2 to 3 cm. In 2021 it reported 10 driverless tandem rollers and two driverless pavers on a section of the Nanjing–Shanghai Expressway, and Sany reported a driverless fleet of pavers and rollers on a concrete paving job. In May 2025 China Daily photographed a driverless paver and six autonomous rollers starting work on a new expressway in Guizhou province. The accuracy and output figures are the makers' own.
Credit
XCMG; Sany
Watch for
Driverless fleet, measured gains
On the Meizhou–Xingning section of the Shantou–Meizhou Expressway upgrade in Guangdong, 10 large machines without drivers have paved and compacted in formation at about 4 km/h since November 2025, started with one command from a tablet. The project's data say the system cut labor costs by 62% and raised overall efficiency by about 50%, with BeiDou satellite positioning held within 3 cm.
Achieved whenthe road owner or an independent laboratory publishes density, smoothness and crew-hour results for driverless and crewed sections of the same job.
Claimed by
the Shantou–Meizhou Expressway project, reported by Nanfang Plus, April 2026
Open
01
Roller pass counts from video
Software that counts how many times the rollers pass over each patch of a fresh asphalt layer, from video taken by a camera on a tripod outside the work area, draws a coverage map, and is checked against independent counts.
The briefCredit · unclaimed02
Roller readings against cores
One open table of every published pair of an intelligent compaction reading and a density test at the same spot, with the roller make, the kind of reading, the material and its temperature, and a test of how well each kind of reading predicts density on projects it was not fitted to.
The briefCredit · unclaimed
Open 01 · the brief
Roller pass counts from video
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
Software that counts how many times the rollers pass over each patch of a fresh asphalt layer, from video taken by a camera on a tripod outside the work area, draws a coverage map, and is checked against independent counts.
Why it is still open
Crews compact to a rolling pattern, a set number of passes, and the operator is usually the only one keeping count. Camera methods so far are prototypes that nobody has checked against a reference. A 2016 West Virginia University thesis by Mohammad Hadi Niki Rashidi mapped passes with an infrared camera mounted on the roller and concluded that the system "should be independently validated." A 2020 follow-up by Fei Dai, John Zaniewski and Linjun Lu used the same kind of roller-mounted camera, noting that fitting an existing roller with intelligent compaction can cost up to $100,000, and judged its accuracy by drift over a long run. In 2021 Gabriel Staub and colleagues at the University of Concepción, Chile, tracked rollers marked with magenta plates in drone video and drew their paths, without counting passes per spot.
researchrepository.wvu.edur3utc.psu.eduisprs-archives.copernicus.org
Built with what exists
a phone or action camera on a tripod; open detection and tracking software such as Ultralytics YOLO with the ByteTrack tracker, or Meta's SAM 2 for following objects through video; OpenCV to map the picture onto the ground from four measured points; Python.
First step
With the owner's and contractor's permission, film 20 minutes of rolling on a parking-lot job from outside the work area, and count passes by hand along one 10 m strip.
Done looks like
Three filmed jobs with a pass count for every 1 m by 1 m cell, checked against counts made by two people from the same video, and against the roller's own satellite log where it has one. Published: the agreement rate, the videos and the labels.
Open 02 · the brief
Roller readings against cores
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
One open table of every published pair of an intelligent compaction reading and a density test at the same spot, with the roller make, the kind of reading, the material and its temperature, and a test of how well each kind of reading predicts density on projects it was not fitted to.
Why it is still open
FHWA's 2011 pooled-fund report found the link between IC readings and asphalt core densities "inconsistent" and called for models that correct for asphalt temperature and separate the layers. Its 2014 follow-up, by George Chang and colleagues at the Transtec Group, studied nine field sites, found that the final readings "do not correlate well with core densities", and fitted a density model using Utah and Maine data. Searches of Crossref and GitHub found single-project and few-site studies, but no pooled public table and no test across projects.
Built with what exists
FHWA and state report PDFs; pdftotext; an AI model that reads tables from scanned pages, such as Claude Fable 5.2 (see the Note "Three working devices"); Python with pandas and scikit-learn.
First step
List the pooled-fund demonstration reports and the 2014 study's appendix tables, and note which print point-by-point data rather than only fitted lines.
Done looks like
At least 1,000 pairs from at least 20 projects in at least five states, including the nine sites of FHWA's 2014 study, each row traced to a report page, and for each kind of reading the error, in percent density, when predicting a project left out of the fit.