Hangzhou launches City Brain with Alibaba Cloud ET

Background

Cities generate a great deal of data about their roads, but the data and the equipment that acts on it have usually been separate systems. Wang Jian, chairman of Alibaba Cloud's technology committee, described the problem in plain terms: traffic cameras and traffic lights frequently stand on the same pole, yet the two are never connected through data ^[2]. A camera may record congestion while the light beside it keeps running on a fixed cycle. Closing that gap — sensing the road and acting on what was sensed within the same loop — was the problem the Hangzhou project set out to address.

The launch

Alibaba Cloud's official press release is dated 13 October 2016 in Hangzhou and was issued at its Hangzhou Computing Conference ^[1]. It announced the Hangzhou City Data Brain project. The project was led by the Hangzhou municipal government and used Alibaba Cloud's artificial intelligence programme "ET" together with big-data analytics to carry out real-time traffic prediction ^[1].

Chinese media framed the announcement in the language of urban history as well as of technology. China News Service described the project as installing an AI hub for a city with more than 2,200 years of history ^[3].

How it was built

The project was not a single-vendor effort. It involved 13 companies including Alibaba Cloud, whose scientists formed a joint research and development team ^[1][3]. Underneath the applications, the computing platform ran on Apsara ("Feitian"), Alibaba Cloud's self-developed large-scale general computing operating system, which the company says can connect million-scale servers into a single supercomputer ^[1]. In practice that means many machines are pooled so that a city-scale data workload can be processed as if it ran on one very large machine.

On top of that platform, the automated traffic system could adjust traffic lights directly — for example extending green time when vehicles changed direction ^[1]. The stated ambition went beyond the signals themselves. According to CNR, the project team planned to digitise energy and water infrastructure and to connect scattered city data resources ^[2]. Traffic was the first application, but the underlying idea was a shared layer through which different parts of a city's operations could be observed and coordinated.

Early results

The first public pilot ran in Xiaoshan District, beginning in September 2016. According to Alibaba Cloud's release, it raised vehicle travel speeds by up to 11% ^[1]. The figure refers to how fast vehicles actually moved on the roads of the pilot area, which is the outcome that signal-timing changes are meant to influence: with better timing, the same road network carries traffic at a higher average speed instead of leaving vehicles waiting at red lights.

Why it mattered

Alibaba Cloud quoted Wang Jian calling City Brain a forward-looking experiment in using AI for social management ^[1]. That framing is worth noting. The project was presented not simply as a traffic-engineering tool but as an attempt to apply artificial intelligence to how a city is run, with traffic as the first testing ground. The combination described at the launch — a municipal government as lead, a group of companies contributing research staff, a general-purpose computing platform underneath, and prediction feeding directly into physical infrastructure such as traffic lights — set out a template for later discussions of city-scale AI. The Xiaoshan pilot gave the idea a measurable early result to point to, and the plan to digitise energy and water systems and connect scattered data resources indicated where the project intended to go next ^[2].