If any of you are watching from outside, thinking China looks far too futuristic, get this: China now has robot police units. They began work over the May Day holiday, patrolling sections of Hangzhou’s (杭州) West Lake (西湖) as a traffic squad.
For now, their job is not to replace human police wholesale. Instead, the robots are taking on the repetitive, high-friction parts of street-level traffic management: guiding tourists, warning pedestrians, directing non-motorised vehicles, and helping manage flows at busy intersections.
At the West Lake – an area swamped with tourists over holiday periods – visitors can press a ‘speak’ button on the robot’s screen and ask for directions. Using speech models, real-time traffic data and geographic information, the robots can recommend walking or public transport routes via voice and on-screen prompts.
At intersections, the robocops get disciplinarian. Equipped with visual recognition algorithms, the robots can identify common violations: e-bike riders without helmets, vehicles stopping beyond the line, pedestrians standing in motor lanes. They don’t have the ability to punish offenders yet, but it’s tempting to imagine a future in which that could be a possibility.
Hangzhou is not alone. In Wuhu, Anhui Province (芜湖市，安徽), a humanoid robot known as Smart Police Unit R001 (芜优智警R001) has already been assisting traffic officers, using cameras, voice alerts and large-model algorithms to spot violations and issue warnings. It can also link with traffic lights and perform standard command gestures in sync with signal changes.
The Dao view: China, robot police, and the push for embodied intelligence
The bigger story is China’s push to move embodied intelligence from showroom demo to street-level infrastructure. State-linked research has projected China’s embodied intelligence market could reach RMB 400 billion (about US $56 billion) by 2030. For now, the robots are still assistants, but those of you feeling like China already looks like a scene from the future likely haven’t seen anything yet.
