Industries / Construction and property / Heavy civil / Paving
Maintenance and repair
Maintenance keeps a pavement in service by sealing its cracks, renewing its surface and grinding it smooth, and by deciding each year which roads get which treatment.
5
Milestones achieved, from 1965 to 2016.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1965
Grinding off the steps
Concrete roads are laid in slabs, and over the years heavy trucks push the slab ends out of line, so each joint becomes a small step, a defect called faulting. A study published by the Portland Cement Association (PCA) in 1999 dates the first use of continuous diamond grinding, cutting the surface flat with closely spaced diamond saw blades, to 1965, on a 19-year-old section of the San Bernardino Freeway (I-10) in southern California that had excessive faulting. The study surveyed 60 sections at 54 sites and put the average life of a diamond-ground surface at 11.4 years, with a 98% chance of lasting at least 8 years before it needed more work.
Credit
California's state highway agency; study by Shreenath Rao, H. Thomas Yu and Michael I. Darter
1970s
Thin coats, open fast
A slurry seal is a cold, runny mix of fine stone, water and asphalt emulsion (asphalt broken into tiny droplets in water) spread a few millimetres thick over a worn road. By a 2013 account in Construction Equipment, the first such wet mix was laid at Berlin-Staaken Airport in Germany in 1936, and microsurfacing, a tougher version, was also first used in Germany, on the autobahn. The Federal Highway Administration (FHWA) describes microsurfacing as a slurry seal with a polymer-modified binder and better materials, laid 10 to 15 mm thick, developed in Europe in the 1970s and introduced in the United States in 1980. It fills wheel ruts and can usually carry traffic about one hour after it is laid.
Credit
German road builders; FHWA names systems sold as Ralumac, Macroseal and Durapave
1970s
Roads scored 0 to 100
PAVER is a pavement management program. Inspectors record each distress in sample areas, such as cracks, potholes and ruts, by type and severity, and the program turns the tally into a pavement condition index (PCI) from 0 (failed) to 100 (excellent) and forecasts the repairs each pavement will need. It was developed in the late 1970s to manage the US military's pavements, at the US Army Corps of Engineers' Construction Engineering Research Laboratories (CERL). ASTM, the standards body, adopted the PCI for airports in 1993 (D5340) and for roads and parking lots in 1999 (D6433), and NATO has adopted it too.
Credit
M. Y. Shahin and colleagues at CERL, with support from the US Air Force, Army and Navy, the Federal Aviation Administration and FHWA
2014
One driver seals the joints
Up to 80% of the surface water that gets into a pavement enters at the joint between the lane and the shoulder, by a study the report cites, and sealing it by hand puts a crew on foot beside traffic. The Sealzall, built by the Advanced Highway Maintenance and Construction Technology (AHMCT) Research Center at the University of California, Davis, with Caltrans, lets one operator seal that joint from the cab while driving at 2 to 5 mph inside a moving lane closure. In a one-year trial, Caltrans District 11 maintenance crews sealed 5 to 8 linear miles of edge joint a day with it. Random cracks inside the lane are still sealed by hand, in a fixed lane closure.
Lessonan earlier joint-sealing machine relied on workers riding in the truck bed to feed sealant blocks into the melting kettle, and a Caltrans rule against riding in a moving vehicle ruled that step out.
Credit
AHMCT (Duane Bennett, Steven A. Velinsky) and Caltrans
2016
Lasers replace the fall survey
Saskatchewan's Ministry of Highways and Infrastructure used to rate its roads each fall by staff checking sample stretches by eye. In 2016 it switched to the Laser Crack Measurement System (LCMS), mounted on an Automated Road Analyzer (ARAN) survey van. LCMS uses laser line projectors and high-speed cameras to record a 3D profile of the lane at up to 100 km/h, and its software picks out cracks, ruts, potholes and ravelling, the loss of stones from the surface. The ministry says the change took people off the road, replaced spot samples with continuous measurement and improved data quality. Its 2014 pilot found that the machine's crack types and severities did not match the visual method, so the ministry first wrote its rules into standards the software could apply.
Lessona machine survey has to be given the owner's definitions before its numbers mean the same thing.
Credit
INO, Canada's National Optics Institute, which developed LCMS; Pavemetrics, which sells it; Saskatchewan Ministry of Highways and Infrastructure
Watch for
A robot that follows cracks
Researchers at Chang'an University in China and Michigan Technological University built APCSbot, a four-wheeled robot that finds cracks with a camera and a neural network and seals along them. From their own tests they report 80.21% accuracy in picking out cracks on three image sets, a sealing speed of about 0.05 m/s and an average sealing error of 5.17 mm on slender cracks.
Achieved whena road agency or an outside laboratory runs it on real pavement cracks and publishes its own count of length sealed and sealing error.
Claimed by
Jian Qi Zhang, Xu Yang, Zhanping You and colleagues, April 2026, in the China Journal of Highway and Transport
A crack-sealing robot for Australia
Charles Darwin University (CDU), Civiltech Solutions and the Additive Manufacturing Cooperative Research Centre signed an AU$1.2 million partnership to build a robot that detects and repairs road cracks with minimal human input, by adding a 3D-printed sealant dispenser to Civiltech's existing LiDAR (laser scanning) platform. CDU's Centre for Asphalt and Road Technologies receives AU$500,000 of it.
Achieved whenthe robot seals cracks on a working road and the road authority reports the length sealed and how the seals held.
Claimed by
CDU and Civiltech Solutions, April 2026
Open
01
A shared crack-sealing robot test
A public test pad and scoring rules for crack-sealing robots: slabs with a known set of real cracks, each traced and measured by hand, and a script that scores every run from overhead photos for the share of each crack filled, how far the sealant strayed from the crack, the sealant used and the time taken. Then run one robot on it 20 times.
The briefCredit · unclaimed02
When fresh seal takes traffic
A small kit of a camera, an infrared thermometer and a humidity sensor, with a model that predicts from the surface's color and temperature and the weather when a fresh microsurfacing or slurry seal is strong enough to carry traffic, checked against a measured test.
The briefCredit · unclaimed
Open 01 · the brief
A shared crack-sealing robot test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test pad and scoring rules for crack-sealing robots: slabs with a known set of real cracks, each traced and measured by hand, and a script that scores every run from overhead photos for the share of each crack filled, how far the sealant strayed from the crack, the sealant used and the time taken. Then run one robot on it 20 times.
Why it is still open
Each team reports its own numbers on its own cracks. APCSbot's 5.17 mm error comes from its makers' tests. Pave Robotics, a start-up, says its Tracer robot seals cracks "faster and more effectively than a crew of six people." Caltrans is evaluating one commercial robotic crack sealer in a project that ran from June 2024 to May 2026. A 2026 Purdue study for the Indiana Department of Transportation counted 51 autonomous maintenance products in eight categories, crack sealing among them, and found "cautious, technology-specific acceptance" among the department's staff. No shared test or published scoring rule was found.
Built with what exists
Slabs of cracked old pavement from a recycler, kept on private ground; a cold-applied crack filler instead of hot-pour sealant, so nothing on the pad is hot; a small gantry or wheeled robot carrying a caulking pump; a phone or depth camera; open crack image sets such as DeepCrack and CFD to train a crack finder; OpenCV; an AI agent that writes the robot's program and rehearses it in simulation, as in Berkeley's agentic robotics work (see the Note "Readable robot programs"), or a robot model taught from a few demonstrations, such as NVIDIA GR00T 2 (see the Note "Five demonstrations per task").
First step
Collect four slabs of cracked old pavement, photograph each from above with a scale bar, and trace every crack by hand to make the first answer key.
Done looks like
A published pad design with at least 20 cracks, each with its path and width measured by hand; 20 filmed runs; a public tally of the share of crack length filled, the mean and worst distance in mm between sealant and crack, grams of filler per metre and minutes per run.
Open 02 · the brief
When fresh seal takes traffic
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A small kit of a camera, an infrared thermometer and a humidity sensor, with a model that predicts from the surface's color and temperature and the weather when a fresh microsurfacing or slurry seal is strong enough to carry traffic, checked against a measured test.
Why it is still open
Opened too early, a new microsurfacing ravels, shedding its stones under tires. Caltrans's maintenance guide measures strength gain only in the laboratory, with the ISSA TB 139 cohesion test of the International Slurry Surfacing Association, and gives a rule of thumb for the field: it "can be opened to traffic when it has turned black." The same guide warns that drivers often drive on it before it has cured. A search for field sensors or instrumented checks of cure time found none.
Built with what exists
Bench samples hand-mixed as in the ISSA laboratory method, from aggregate, a polymer-modified asphalt emulsion, cement and water, cast in molds; a phone camera and a color card; a low-cost infrared thermometer; a temperature and humidity logger; a cohesion tester, or a home-made scuff test with a weighted wheel turned in place; Python with scikit-learn.
First step
Ask an emulsion supplier for a sample of microsurfacing emulsion, mix 10 bench samples, and photograph them every 5 minutes as they darken.
Done looks like
60 samples cured at three or more temperatures and humidities, each scuff- or cohesion-tested every 10 minutes until it passes. Published: the kit's predicted ready time against the measured time, the mean error in minutes, and the same error for the "turned black" rule.