How Hangzhou’s 15 T2 Traffic Robots Are Changing Routine Road Management

Hangzhou’s much-publicized “robot police” squad is not an autonomous law-enforcement unit. Since May 1, 2026, 15 wheeled humanoid T2 robots developed by SUPCON Information have been working at key intersections around West Lake and in major commercial areas, where they guide traffic, assist tourists and remind people about routine road rules. Human officers remain responsible for formal enforcement.

What the T2 robots do

The T2 stands 1.88 meters tall and weighs about 98 kilograms. Its cameras, radar and onboard computing are used to monitor road scenes and identify behaviors such as:

• riding an e-bike without a helmet;
• crossing or stopping beyond a stop line; and
• violating traffic signals.

The robots can issue spoken or displayed reminders, provide directions and use articulated arm movements modeled on standardized traffic-control gestures. Some reports also describe the units as assisting with parking or connecting people to police services in emergencies.

Their humanoid form is functional as well as theatrical: a visible head, torso and moving arms make the machines easier for pedestrians and drivers to notice than an ordinary roadside sensor. But the deployment remains geographically limited, with units assigned to designated intersections and routes rather than given unrestricted citywide freedom.

A daytime, human-monitored operating model

Available reporting commonly places the robots’ regular operating schedule at about 7 a.m. to 6 p.m. They work along preset routes or at predefined locations and are monitored by police teams through a central platform.

That makes the program closer to supervised robotic traffic assistance than to independent patrol. The machines can sense, signal, speak and record, but a human-controlled system remains in the loop when an incident needs to be reviewed or escalated.

The reported results are promising—but inconsistent

SUPCON has said the fleet issued more than 170,000 warnings and helped reduce relevant monthly violations by more than 40 percent. Those figures were reported without independent verification; Reuters noted that Hangzhou traffic police did not respond to a request for comment.

Earlier local reporting gives a different picture. People’s Daily reported more than 25,000 traffic-violation warnings and 647.7 hours of service after the squad began operating, while other coverage cited 11,897 warnings during the first three days of the May holiday.

The figures may reflect different periods, categories or counting methods, but the supplied reporting does not reconcile them. Nor does it provide a controlled comparison showing how much of the reported reduction came from the robots rather than seasonal traffic patterns, conventional police activity or changes in crowd levels. The results should therefore be treated as operational claims, not proof that the robots caused a 40 percent improvement.

Why the robots cannot fine, stop or detain people

Despite the “robocop” label, the T2 units do not have the authority to make arrests or issue fines. Their role is limited to reminders, traffic direction, sensing, recording and referral to human officers.

This boundary is central to the pilot’s design. A robot can flag a possible violation, but a human official must remain responsible for deciding what happened, assessing the evidence and determining whether formal action is appropriate. Keeping penalties and coercive interventions with people also creates a clearer line of institutional accountability when a machine misidentifies an event or behaves unsafely.

In practical terms, the robots are intended to remove repetitive roadside work from officers—not to replace the judgment involved in enforcement.

Accuracy and bad-weather performance remain open questions

The available reports describe the T2’s sensors and target behaviors, but do not provide independently validated accuracy rates for helmet detection, stop-line detection, signal violations or identity matching. They also do not establish a detailed weather-performance specification.

Cameras and radar can provide complementary information, but real intersections are difficult environments. Rain, glare, fog, crowds, occlusion, reflective surfaces and worn lane markings can all make detection harder. Claims that the robots can continue operating through poor weather should not be interpreted as a published guarantee of consistent accuracy across those conditions.

For a city considering procurement, the missing metrics matter as much as the headline capabilities. Useful evaluation would include false-positive and false-negative rates, uptime, intervention frequency, response to blocked sensors and the quality of evidence handed to officers.

Human officers still carry the operational risk

The human-machine arrangement does not eliminate the need for police. Officers must supervise the fleet, validate recorded footage, handle disputes, respond to emergencies and decide whether a warning should become formal enforcement.

They also remain the natural point of accountability for unsafe interactions or erroneous penalties. That makes the robots potentially useful for standardized, high-volume tasks while preserving human control over situations involving ambiguity, conflict or legal consequences.

What SUPCON is really testing

SUPCON’s stated objective is to transfer repetitive traffic-control duties to machines so officers can focus on work requiring judgment and intervention.

That makes Hangzhou a significant commercial test, not simply a technology demonstration. A public intersection exposes the robots to unpredictable pedestrians, vehicles, weather, maintenance demands and public scrutiny. It also tests whether a humanoid platform can deliver a recurring municipal service: traffic guidance, visitor assistance, basic compliance reminders and incident reporting.

The business case will depend on whether those services justify the full cost of procurement, supervision, maintenance, connectivity, software updates and integration with existing traffic and police systems.

How widely is the technology spreading?

The clearly documented Hangzhou deployment consists of 15 T2 robots and has been described as China’s first formally organized robot traffic-management squad.

One later report says SUPCON told Reuters that nearly 50 robots were operating across roughly eight cities. That suggests possible expansion beyond Hangzhou, but the supplied reporting does not establish the size, duration or official status of each deployment. It is therefore too early to describe T2 robots as broadly deployed across China.

There is also no solid evidence in the supplied sources of signed large-scale municipal orders or overseas T2 deployments. Expansion would require cities to demonstrate reliable performance and connect the robots to traffic lights, dispatch systems, video platforms and police information technology.

The regulatory and export challenge

Domestic and overseas deployments would face different combinations of public-space robot-safety rules, communications approvals, cybersecurity requirements, evidence-handling standards and privacy protections for video and location data. Systems that process identifiable or biometric information would face additional scrutiny.

Hangzhou’s regulatory framework is also evolving. A report on the city’s embodied-AI-robot regulation says it took effect on May 1, 2026, covering areas including industrial management, applications and security. That indicates the legal environment is developing alongside the technology rather than being fully settled.

For export markets, the hardest questions may be less about whether a robot can raise its arm and more about who is legally responsible for its decisions, how incident footage can be used, how the system is certified for public roads and whether it can interoperate with local infrastructure.

The broader lesson

Hangzhou’s T2 squad is not evidence that robots are replacing traffic police. It is evidence of a more limited and commercially practical experiment: whether humanoid robots can perform visible, repetitive and supervised municipal work in live public settings.

The pilot’s success will be measured by more than warning counts. Cities will need credible evidence on accuracy, safety, uptime, total cost and privacy. If those numbers work, the T2 could become a model for narrow human-machine collaboration. If they do not, its most lasting contribution may be showing how difficult it is to turn a striking robotics demonstration into dependable public infrastructure.