The best evidence anyone has for a traffic technology installed on an American freeway came from switching it off. In the autumn of 2000 the Minnesota Department of Transportation turned all 430 ramp meters in the Twin Cities dark and left them dark for six weeks, because the state legislature had ordered a study and the only way to measure what the meters were doing was to stop doing it.
Cambridge Systematics published the results in 2001, with the final report issued 13 May 2002. Freeway volume fell 9 percent. Travel times rose 22 percent. Speeds fell 7 percent. Crashes rose 26 percent once the figures were averaged for seasonal variation, with side-swipe collisions up 200 percent and run-off-the-road crashes up 60 percent. The state’s answer in 2002 was not to restore the old timings but to install a responsive system with wait limits: no more than four minutes on local ramps, no more than two minutes on freeway-to-freeway ramps.
The public verdict was more interesting than the traffic counts. Most survey respondents believed conditions had worsened while the meters were off, and over the same period support for modifying the metering system rose from 60 to 70 percent, with the modifications people named being faster cycle times, shorter operating hours and fewer meters. Drivers judged the system effective and wanted less of it. Every smart highway technology since has had to live inside that tension.
The deployed layer is thin, old, and unevenly evidenced
Metering is the senior technology. Variable speed limits are the next one down, and the federal evaluation of them is a lesson in how much a single acronym can hide. FHWA’s Turner-Fairbank centre published crash modification factors for variable speed limits in May 2021, using data from three corridors in Virginia, Wyoming and Georgia.
The three results do not agree, and the report explains why rather than averaging them away. Virginia’s system exists to manage speeds in fog, and the research team estimated the displayed limit sat constant at 65 mph roughly 95 percent of the time; no crash shift at that site reached statistical significance, and the estimated factor for total crashes was 1.23, pointing the wrong way. Wyoming’s rural corridor, where the system manages operations continuously rather than waiting for weather, produced significant reductions of 34.4 percent in total crashes and 65.2 percent in rear-end crashes, a figure the authors themselves flagged as a very large estimate. Georgia was the only site with comparison corridors, which is why the team called that analysis its most robust; there the significant reductions were 29.2 percent total and 35.2 percent rear-end, smaller in general than Wyoming’s, with day crashes reaching significance only at the 90 percent level. Benefit-cost ratios came out at 40.38 for Georgia and 9.05 for Wyoming.
The conclusions carry the sentence that should govern how those numbers are used. The study compared safety performance without accounting for differences in management strategy between agencies, so its findings quantify a measurable change after installation and say nothing about the relative contribution of the agency’s operating approach, the algorithm that chooses the displayed speed, or the level of congestion. A crash modification factor for variable speed limits is a factor for a bundle of sign hardware, sensor spacing, control logic and institutional habit, and the bundle differs at every site.
The most ambitious version was measured, and one site came out at roughly break-even
Integrated corridor management is the closest thing the field has to a real smart corridor: freeway, arterials and transit operated as one system, with a decision support tool proposing response plans when something goes wrong. FHWA funded two full deployments, 28 miles of U.S. 75 in Dallas and I-15 in San Diego, and published an independent evaluation of both.
The honest part comes first. A before-and-after analysis using field data was abandoned for want of sufficient system activations, so the mobility results rest on post-deployment modelling and simulation, using different tools at each site, which the evaluators noted makes direct comparison inappropriate. San Diego’s benefit-cost range came out at 2:1 to 9:1 across 17 response plan implementations. Dallas came out at 0.55:1 to 1.64:1 across 35. The flagship American smart-corridor program produced, at one of its two demonstration sites, a range that straddles break-even.
What the evaluators found unambiguously was institutional. Interagency cooperation was a big success at both sites, operators reported better situational awareness, and both corridors ended with substantially more message signs posting travel times than they started with. Thirteen further sites received planning grants of up to $200,000, and FHWA’s own outlook concluded that incremental deployment is the most likely viable path forward, with corridor management competing for money against conventional projects and many grant recipients unsure what a finished system would even look like. An outlook that plain is worth more than a press release. The question of how these projects reach a capital program at all runs through how highway projects are planned.
The frontier, priced by the quarter mile
In October 2024 an electric semitrailer drove a quarter-mile testbed on U.S. 52 and U.S. 231 in West Lafayette at 65 mph and drew 190 kilowatts from coils buried in the concrete beneath it. Indiana’s transportation department, Purdue University and their partners had been working toward that run since 2018, and the department’s commissioner, Lyndsay Quist, said the test demonstrates technology that could help lower the costs of building electrified highways.
The engineering is real and the arithmetic is sobering. A quarter mile at 65 mph takes about 13.9 seconds, so 190 kilowatts delivers roughly three quarters of a kilowatt hour per pass. Dynamic wireless charging is therefore not a station technology at all. Its unit of account is energised lane miles, which means its cost per unit of useful energy falls only as the electrified fraction of a route rises. Indiana’s published sequence respects that: laboratory and pavement optimisation first, then a quarter-mile testbed aimed at 200 kilowatts and above for heavy trucks, then an interstate segment at a location still undetermined, each stage conditional on the one before. The comparison with plaza charging is taken up in how EV charging will change interstate highways.
Smart pavement covers two different propositions that are usually sold as one. Indiana’s magnetizable concrete is the version where the pavement is the device, and it inherits the pavement’s problems: the coils have to survive the same freeze cycles, the same studded tires and the same overlays as the surface course above them. The sensing version, where fiber and electronics ride inside the slab to report vehicle position and surface condition, asks a structure designed for a multi-decade service life to carry components with a much shorter one, and that is a maintenance question before it is a data question. The service-life arithmetic that decides it sits in long-life pavement design, and the broader technology picture runs through the technology section.
What survived measurement has something in common
Metering restrains traffic. Variable speed limits slow it. Corridor management informs it and reroutes it. None of the deployed technologies with published evaluations adds capacity, and the two whose benefits are least contested are the two that take something away from the driver, which is why both provoked the political fights they did.
The frontier is being assessed on capability instead: a truck charged at speed, a pavement that reported a departure from the lane. Those are demonstrations of function, and the Minnesota experiment set a different and higher standard, which is to install a system at scale, remove it, and count what changes. Nothing on the frontier has been subjected to that standard, and nothing will be until a deployment is large enough that switching it off would be felt.