Wuhan is Baidu's single biggest robotaxi market and, for a few tense weeks in 2026, the reason the whole Chinese industry paused: a cloud outage froze more than a hundred driverless Apollo Go cars at once, and the fallout reached every other city testing autonomous vehicles nationwide. That headline sits on steadier ground — a fully driverless metro line running since 2021, a national vehicle-testing zone wiring roadside cameras and radar into nearly a quarter of the city since 2019, and a citywide hazard-sensing platform switched on in December 2024 with 15,000 sensors on bridges and tunnels alone. The pattern is transport automation and disaster sensing, both built fast and mostly self-reported so far; Wuhan sells its public data through a state-run marketplace rather than opening it, which is a genuinely different choice from most peer cities, not a lesser one.
The atlas documents 5 projects in Wuhan, the earliest dating to 2019: 5 still running without a verdict. Every entry carries an evidence grade — the strongest here is official (4 entries) — and links its sources.
Three entries that show what Wuhan is actually doing.
Computed from the 5 entries below — a profile, not a score.
The thing that gets the headlines in Wuhan rides on top of something slower and harder to build. In September 2019 the city's development zone opened the national intelligent connected vehicle demonstration zone, China's sixth and the first in the centre of the country, issuing what it says were the world's first commercial autonomous-vehicle permits on the same day. What it has actually been building since is roadside infrastructure: from 191 kilometres of vehicle-to-road cooperative roads and a closed test track at launch to 3,487 kilometres of open smart roads by April 2025, close to a quarter of the city's area, with masts carrying cameras, millimetre-wave radar and lidar into edge-computing units the zone says detect obstacles within 200 metres. In 2025 it recruited 10,000 ordinary residents to ride in connected traffic as what it calls chief experience officers. Six years of incremental, road-by-road build-out is the asset most coverage skips past.
The service on top of it made Wuhan famous and then made it a cautionary tale. Apollo Go began carrying passengers with a safety driver in May 2022 and by that August made Wuhan the first Chinese city with fully driverless commercial rides, across thirteen square kilometres of the development zone; it is now the operator's largest single-city deployment, claimed at more than 3,000 square kilometres — while the city's own transport bureau put the active fleet at 400 vehicles in 2025 and called the widely repeated claim of more than a thousand a rumour. Fares undercutting taxis by roughly 60 to 75 per cent produced a genuine driver-livelihood backlash. Then on 31 March 2026 a cloud-side fleet-management outage froze more than a hundred cars simultaneously across the city for up to two hours, with in-car emergency call systems silent and passengers stranded. A nationwide freeze on new autonomous-driving permits followed from 29 April, lifted only after an industry-wide safety review in late June, with vehicles — some carrying safety drivers again — back on Wuhan's streets from early July.
In late 2024 the municipal government built the densest monitoring layer in the record. The city safety risk platform went live on 25 December, meant to replace after-the-fact emergency response with prediction across lifeline infrastructure, public safety, workplace safety and natural disasters: it pools weather, drainage and river-level data, watches 200 mapped waterlogging hotspots, and reads 15,000 sensors and 3,200 cameras across roughly 700 bridges and more than a hundred tunnels, on a digital-twin base map encoding over thirty million objects, routing alerts through a municipal, district and community chain now being extended across the province. Three months earlier a single district had switched on its own operations dashboard, whose distinctive layer is nine automated drone stations that had logged more than 14,500 cumulative flight hours and reported 235 incidents by the launch date — river and lake patrols, forest-fire watch, illegal land use, construction-waste vehicles.
The newest entry inverts what most cities in this atlas do with public data. In October 2025 the city's state-owned data group began piloting HanData, a marketplace that licenses processed public data to companies under contract rather than publishing it as an open download, and within roughly four months logged more than 70,000 data-product transactions across 1,252 listed products and 707 registered entities. Its clearest use case is insurers verifying an applicant's health history in seconds instead of three to five days; a separate feed streams live signal timing from 640 intersections into countdown displays and green-wave routing. In March 2026 the marketplace was connected to China's national data infrastructure. It is closer in spirit to a state-run data exchange than to a transparency platform, and every figure so far is the operator's own.
On a time axis, by the year each one started — 2019 to 2025. Pick a dot to read what it was.
✔ Every published entry is checked against the public sources it cites before publication.
Editorial links between Wuhan's own entries — the city as a system, not a list.
Plans and announced trajectories — not yet documented results.
Wuhan's most important entry is the one that failed, and it failed in a way the rest of this atlas has not yet recorded elsewhere: not a sensor, not a driving decision, but a cloud-side fleet-management outage that froze more than a hundred cars across a city at once and paused an entire national industry for two months. The question that leaves open is whether anything published since names what the recovery path is when the link to the fleet drops, because the in-car emergency systems went silent at the same time. Two others follow the city's own pattern. The roadside network now covers close to a quarter of the city and its accuracy figure is self-reported, so the natural next step is whether anyone outside measures what a 3,487-kilometre sensing layer detects. And the data marketplace is the mirror image of an open-data portal: it prices what other cities give away, which makes its transaction count a real signal of demand and its absence of an independent audit a real gap.
One city the atlas reads next to Wuhan — the reason first, then the entries it rests on. Chosen by hand from the Wuhan threads below, not counted out of them.
Two Chinese cities running the same national programme in parallel — robotaxis from one operator, a wired-road pilot zone, bridges under sensors — and both report kilometres of instrumented road rather than journeys served.
Every editorial link from Wuhan's entries into another city, strongest first. Each one is a statement about two projects, not about two cities — the three above are the ones the atlas is willing to make at city level.
1001 Smart Cities (2026). “Smart-city projects in Wuhan, China.” The Smart City Atlas. https://1001smartcities.org/cities/wuhan/ (last verified 2026-08-02). Data: CC BY 4.0.
The underlying data is free to reuse with attribution — see the open data page.