Tech • AI • Robotics
New AI Robot Police Are Taking Over China’s Streets
7/10TL;DR
Hangzhou is deploying autonomous traffic robots that issue spoken safety warnings rather than fines, as China simultaneously pushes robots into increasingly demanding real-world public and athletic roles.
KEY POINTS
Traffic robot on active street duty
A wheeled robot called the T2, built by Subcon Information in Hangzhou, is operating at city intersections rather than in a lab or trade show setting. Standing about 6 feet 2 inches tall and weighing 98 kilograms, it patrols on wheels and works full daytime shifts.
What the T2 monitors
The robot uses cameras, radar and onboard computing to watch for four behaviors: e-bike riders without helmets, unauthorized passengers on e-bikes, vehicles crossing the stop line, and pedestrians crossing against the signal. When it detects a violation, it gives a polite warning through its speaker system instead of issuing a ticket or fine.
Designed to mimic traffic police gestures
The T2 has seven-jointed arms that reproduce standard traffic police hand signals in sync with the light cycle. It also serves as a roadside information point, allowing people to ask for directions, parking guidance or traffic-rule explanations, and it can connect users to local police in an emergency.
Scale of deployment and claimed impact
Subcon says 15 of the robots have been deployed in Hangzhou since May and have issued more than 170,000 warnings. The company also claims monthly cases of helmetless riding and stop-line violations have fallen by more than 40%, although those figures were not independently verified and local traffic police did not comment publicly.
Operational model and technical limits
The robots reportedly run autonomously from 7:00 a.m. to 6:00 p.m., moving among preset positions without remote driving. Police teams can monitor and dispatch them through a central platform. The company says testing has covered tourist areas, school zones, business districts, rush hour and bad weather, while acknowledging that strong sunlight, rain and temperature extremes can still affect sensing performance. It places recognition accuracy above 95%.
A prevention-first enforcement philosophy
The system is framed less as punishment and more as a repetitive reminder tool aimed at changing habits before unsafe behavior becomes routine. Its core selling point is consistency: unlike a human officer, the machine can repeat the same warning hundreds of times a day without fatigue.
Expansion beyond a single city
Subcon says nearly 50 such robots are now operating across about eight cities in three Chinese provinces, with expectations of approaching 200 units by year’s end. That suggests a broader rollout rather than a small pilot program.
Overseas ambitions face regulatory barriers
Other Chinese robotics groups, including A Moga, backed by automaker Chery, see potential demand in regions where officers work in extreme heat or long rainy seasons, such as the Middle East and Southeast Asia. But exporting police-style robots would require integration with local law-enforcement systems, certifications and compliance with data-protection rules, a major hurdle for machines built around continuous public sensing.
Robotics push extends to athletic competition
The broader backdrop is a rapid acceleration of robotics testing in public and dynamic settings. At the second World Humanoid Robot Games in Beijing, organizers said 2,056 machines from 666 teams across 16 countries took part, with participation up 138% from the previous year.
Tennis and sprinting showcase advanced control
Robotics company Galbot demonstrated humanoids rallying in what was described as the first live autonomous humanoid robot tennis match, completing more than 100 consecutive rallies and recovering from balance losses mid-play. The system uses a method called LATENT, which learns reusable motion building blocks from about five hours of imperfect amateur tennis motion data. In sprinting, Honor’s humanoid Lightning reportedly ran 100 meters in 9.32 seconds, faster than Usain Bolt’s 9.58 world record, while another robot, Tiangong Ultra, posted 9.39. The feats highlighted both gains in speed and the remaining instability of bipedal machines after footage also showed one robot crashing into a barrier.
CONCLUSION
China’s latest robotics deployments show a shift from controlled demonstrations toward machines performing repetitive public-safety and athletic tasks in unpredictable environments. The key question is no longer whether the hardware can move, but whether these systems can be trusted, regulated and integrated at scale.
Explain this