Hangzhou’s T2 Robot Traffic Police Can Warn Drivers—but Not Arrest or Fine Them

Hangzhou’s T2 “robot traffic police” are humanoid-looking, wheeled machines developed by SUPCON Information. Their job is to monitor intersections, guide traffic, answer basic public questions and remind people about minor violations—not to replace police officers or make autonomous enforcement decisions.

The distinction matters. The T2 can identify and report behavior, but the authority to impose a legal penalty remains with human institutions.

What the T2 robots do in Hangzhou

Hangzhou deployed a 15-robot traffic-management squad around the West Lake area on May 1, 2026. The machines operate at busy intersections and other high-traffic locations, where they can coordinate their movements with traffic signals and use standardized arm gestures to direct drivers, cyclists and pedestrians.

A T2 is approximately 1.88 meters tall and weighs 98 kilograms. Despite its humanoid upper body, it travels on wheels rather than legs. Its reported equipment includes cameras, radar and onboard computing for monitoring traffic conditions and recognizing selected behaviors.

Beyond traffic control, the robots can provide directions and other basic assistance to visitors. That makes the system a hybrid of traffic sensor, public-information kiosk and visible roadside reminder platform rather than a conventional patrol officer.

Which violations can the robots detect?

Public reporting identifies several target behaviors:

• Riding an e-bike or other non-motorized vehicle without a helmet
• Carrying passengers in situations where e-bike passenger riding is prohibited
• Stopping or crossing beyond a marked stop line
• Pedestrians crossing against a signal
• Pedestrians entering or remaining in motor-vehicle lanes

When the system detects a suspected violation, it issues a spoken warning. Reports from Chinese state media say that if a person does not correct the behavior after three reminders, the system can archive relevant clues and send them to an early-warning center under Hangzhou’s public-security traffic-management authorities.

That process is better described as detection, warning and referral. It does not mean the robot itself has issued a ticket or established legal liability.

What results has SUPCON reported?

SUPCON says the Hangzhou fleet had issued more than 170,000 warnings since its May launch and was associated with a reduction of more than 40% in monthly helmetless-riding and stop-line violations.

Those figures indicate substantial operational activity, but they should be treated as company-reported results. The available reporting does not provide an independent evaluation, a comparison with similar intersections without robots or a method for separating the robots’ effect from other traffic-safety measures. A high warning count also does not, by itself, prove that the machines caused the reported decline.

Earlier coverage cited 11,897 warnings over three days, or roughly one warning every one minute and 43 seconds. That figure was attributed to Hangzhou Daily reporting and illustrates the system’s capacity for continuous reminders, but it is not an independent measure of accuracy or effectiveness.

Why can’t the T2 stop, fine or detain people?

The robots have no reported arrest authority, weapon or physical-coercion role. Their design is intentionally limited to directing traffic, issuing reminders and passing information to human authorities.

The legal reason is just as important as the technical one. China’s Administrative Penalty Law defines administrative penalties as actions taken by administrative organs under legal authority and procedure. It also states that administrative penalties are enforced by qualified law-enforcement personnel.

A commercial robot can therefore support observation and evidence collection, but it is not itself the authorized legal decision-maker for a fine, arrest or deprivation of liberty. Any later enforcement action must be attributable to the appropriate human-led authority and follow the applicable process.

The system’s practical technical limits

The T2 is designed for structured environments: marked intersections, known traffic signals and a relatively narrow set of observable behaviors. Its reported capabilities should not be confused with general-purpose autonomy.

Several limitations follow from that design:

• Mobility: A wheeled platform is suited to fixed or prepared public spaces but is less adaptable than a human officer in crowded, damaged or obstructed environments.
• Perception: Camera and radar systems must interpret movement, road markings, traffic signals and interactions among multiple road users. Public reports do not establish the system’s accuracy or false-positive rate.
• Escalation: A robot can repeat a warning, but it cannot safely resolve a confrontation, pursue a person or make a judgment about rapidly changing danger.
• Evidence quality: The available reporting does not describe independent testing of collision safety, cybersecurity, data retention, privacy, bias or the evidentiary standards applied to recorded violations.

The evidence is therefore strongest for a narrow claim: the T2 can perform repetitive monitoring and reminder tasks in selected traffic settings. It is not enough to conclude that it can reliably handle the full complexity of policing.

What safeguards are visible—and what remains unclear?

The clearest safeguard is the separation between machine assistance and legal enforcement. The robot may observe, warn and forward information, while a lawful penalty remains the responsibility of an authorized enforcement body. China’s administrative-law framework also provides procedures and review mechanisms for enforcement actions such as fines and other compulsory measures.

For wider deployment, responsible safeguards would need to address at least five areas:

1. Human review before sanctions: A machine-generated alert should not automatically become a penalty without an accountable authority reviewing it.
2. Evidence integrity: Authorities would need rules for preserving, verifying and challenging video or sensor records.
3. Notice and appeal: People should know what conduct was recorded and how to contest an enforcement decision.
4. Privacy and retention: Operators would need clear limits on identifying, storing and sharing footage from public spaces.
5. Safety and accountability: A named operator or agency should be responsible for malfunctions, collisions, inaccurate warnings and system updates.

The available reports do not identify a Hangzhou rule that publicly establishes all of these robot-specific protections. That does not prove that no internal procedures exist; it means the published evidence is insufficient to assess them.

How far could the system expand?

SUPCON has said that nearly 50 similar robots were operating in about eight cities across three Chinese provinces, with deployments expected to approach 200 by the end of the year.

That points to an early domestic rollout, but the figures remain company-reported. The company has also expressed ambitions for international expansion, yet the available evidence does not confirm a particular overseas deployment or export scale.

Overseas use would not be a simple equipment sale. Each market would have to address who may exercise police powers, how public surveillance is governed, what procurement rules apply and how automated decisions are reviewed. Those requirements could make a warning-only traffic-assistance model easier to export than a system connected directly to fines or other coercive powers.

What the T2 says about China’s humanoid-robot industry

The most significant lesson is not that robots are replacing traffic police. It is that companies are finding commercial value in constrained, visible tasks that can be piloted by municipal authorities.

Traffic management offers a useful entry point: the environment is structured, the behaviors are repetitive, the benefits are easy to demonstrate and the machine can remain legally subordinate to human officers. A wheeled design also avoids many of the balance, navigation and safety challenges associated with fully legged humanoids.

In that sense, the T2 represents a pragmatic commercialization strategy. Deploy a machine in a real public setting, gather operating experience, demonstrate a measurable service function and use the pilot to support wider adoption. The program shows how humanoid-looking robots can enter everyday infrastructure without first needing to become general-purpose workers—or autonomous police.

For now, the strongest conclusion is a cautious one: Hangzhou’s T2 robots are useful traffic assistants with meaningful limits. Their reported scale and warning totals suggest an ambitious pilot, while the lack of independent performance and safeguard data leaves open the harder questions about accuracy, accountability and public acceptance.