China Put 15 Robot Traffic Officers on the Street
Fifteen wheeled robots are directing traffic and warning rule-breakers in Hangzhou. Their maker says they issued more than 170,000 warnings and cut two violations by over 40 percent, but police did not confirm those figures and the machines cannot fine or arrest anyone.
At busy intersections in Hangzhou, China, a traffic officer now stands 1.88 metres tall, weighs 98 kilograms and rolls between assigned positions. It is a T2 robot built by SUPCON Information, one of 15 machines that have worked alongside human police since May.
The 30-second summary
- What happened? Hangzhou deployed 15 wheeled robots to direct traffic, answer questions and warn pedestrians, cyclists and drivers about visible violations.
- Why does it matter? The pilot moves humanoid robots from trade-show demonstrations into a public service job with long shifts and real street conditions.
- What is the catch? The strongest performance numbers come from SUPCON, were not independently verified, and do not show how often the robots made mistakes.
KEY NUMBER
SUPCON says the 15 Hangzhou robots issued more than 170,000 warnings since May, a figure that local police did not confirm to Reuters.
What the robots actually do
Cameras, radar and onboard computers look for helmetless e-bike riders, illegal passengers, vehicles over a stop line and pedestrians crossing against a signal. Seven-jointed arms reproduce standard police gestures in time with traffic lights, while a screen and voice system provide directions, parking information and traffic rules.
The units work autonomously between preset positions from 7 a.m. to 6 p.m., according to SUPCON. A central platform lets police teams watch and dispatch them, and an emergency feature connects a member of the public to a human officer.
Why this deployment is more interesting than a robot dance
Many humanoid demonstrations prove that a machine can jump, wave or carry an object for a few minutes. Hangzhou is testing a narrower but more useful question: can a robot repeat a visible, structured task for a full shift among rain, glare, crowds and unpredictable road users?
This is also a test of human-machine boundaries. The robots can observe, remind and forward records, but humans keep coercive powers. That division matters because a false camera match is inconvenient when it triggers a warning and much more serious when it triggers a fine or detention.
Before we overstate the result
- SUPCON says helmetless riding and vehicles crossing stop lines fell by more than 40 percent per month, but Reuters could not independently verify the claim.
- Hangzhou police did not say how many robot referrals led to enforcement, how many detections were false, or how performance varied by location.
- The claimed recognition accuracy above 95 percent comes from the manufacturer. Strong sunlight, rain and temperature extremes can still affect the cameras and radar.
The pilot therefore demonstrates sustained deployment, not proven superiority over human traffic officers. It also raises unresolved questions about retention of street video, access to recorded data and how citizens can contest an incorrect alert.
What happens next
SUPCON says nearly 50 traffic robots are operating across about eight Chinese cities and expects the number to approach 200 by the end of 2026. Expansion will show whether the system works outside tourist districts and whether authorities publish independent error and safety data.
The machines fit a broader effort to make robots useful in ordinary work. NewTqnia has also covered an excavator controller designed around the machine's own movements and an open system for moving robot software between different bodies.
Takeaway
Hangzhou's robots are street equipment, not autonomous police. The meaningful milestone is an 11-hour public deployment with human oversight; the unanswered question is whether independent audits will support the maker's claims about accuracy and fewer violations.
Verified topics and entities
Sources and citations4 sources
External references used to support the reporting in this article.
Published by
NewTqnia Robotics Desk
An institutional editorial team within NewTqnia
