South Korea

What the record shows

South Korea builds automation earlier than almost anyone and then spends years deciding to trust it — and unusually, its own operators document the wait. Incheon Metro Line 2 opened in July 2016 with cabinless trains and signalling designed for unattended running. It ran with safety officers aboard for nearly eight years, after an opening-period derailment that was falsely reported as a drill, and only switched the full line to genuinely unattended operation on 1 January 2024. The number the operator put behind that decision is the honest part: manual interventions fell from 139 in 2017 to one in 2023. A rating is a licence; a licence is not a practice.

The same caution shows at the quay. Korea's first fully automated container terminal opened at Busan New Port in April 2024, where yard cranes and driverless carriers work unsupervised — and the nine ship-to-shore cranes are still remote-controlled by staff, because wind and swell move the ship. That entry also carries the finding most worth exporting from Korea: automation and job cuts are separable. The union settlement held because an entire existing crew was moved across from an older berth and retrained rather than replaced.

Where the record turns uncomfortable is around designations. Korea likes to award a special status and count that as progress, and this atlas holds three cases where the status arrived and the service did not. Busan's blockchain regulation-free zone, the country's only one, secured three changes to national rules and then stopped: tenant companies stood at seventeen for four years, no foreign direct investment arrived, and no new demonstration was approved after 2022. Busan's autonomous-driving pilot zone was designated in November 2022 and produced its first bus anyone could board in September 2025 — a genuine service, now charging the standard fare, and the ministry's own December 2025 review still graded the zone lowest of those it assessed. In Incheon a government-designated test district received the same lowest grade with no public autonomous service running there at all.

The most expensive entry in the Korean record is not a technology at all but a forecast. The Busan–Gimhae light rail opened in 2011 as a private concession with a minimum revenue guarantee written against a consultant's demand study that predicted 176,358 riders a day. Just over thirty thousand came. The two cities paid 906.8 billion won between 2011 and 2025 and are contractually liable for at least 1.09 trillion more through 2041; a 2017 renegotiation swapped the revenue guarantee for cost compensation and trimmed the remaining bill, but could not undo the years already committed. Nearby, Line 4 makes the smaller version of the same point: Korea's first driverless metro is also its least-used line, and because the automation is total, a single fault stops the whole line for hours.

Set against that is a genuine national strength, and it is the meter. Three cities in this atlas are converting their water networks to remote reading — Seoul, Incheon and Busan — and the results separate cleanly by how the rollout was sequenced. Seoul's pilot districts saw meter-freeze incidents fall from 1,853 to 47 across five winters, which is what a hourly-reporting meter is actually for in a city that freezes. Incheon deliberately took the high-loss and hard-to-read areas first, so leak detection arrived before coverage did. Busan is the cautionary one: seven years after its pilot, 27.7 per cent of connections are converted, ranging from 98.4 per cent in one district to 6.3 in another, and a councillor challenged the radio standard the city had picked at pilot stage.

Seoul, meanwhile, is the country's showcase and does two things few cities anywhere do. It designed a public service from movement data rather than a plan: the Owl Bus night network was routed by analysing over three billion anonymised call records and late-night taxi trips, and the city's own evaluation credits it with a measurable fall in taxi refusals and a rise in women's night-time activity. And it publishes the flaws in its own sensors. S-DoT, a network of roughly 1,100 street stations reading seventeen environmental variables every two minutes, is documented by a peer-reviewed quality study that found the readings run systematically warm against the official reference station and that data loss is substantial. That is a city letting an outside team grade its instruments in public, which is rarer than the instruments.

The rest of the Korean record is the physical layer done carefully: food waste charged by the kilogram at card-tapped bins since 2013, streetlight, signal, camera and Wi-Fi folded into one pole to cut duplicate installation cost, a 3D twin of the whole 605-square-kilometre city, and sixteen district control centres in Busan that a headline would happily call one. Read together with Songdo — the built-from-scratch showcase still short of its population target two decades on — the Korean pattern is a country that installs early, waits long, and is more willing than most to write down how long it waited.

Sources beyond the atlas

Every factual claim above is carried by one of the linked atlas entries; no external source was needed for this section.