On 21 October the Cyberspace Administration of China's news feed carried a report from Tongxiang, Zhejiang, on a robot traffic officer stationed at the Zhongxing Road and Zijin Road junction in Puyuan, the second venue for the 2026 World Internet Conference Wuzhen summit. The unit is 188 cm tall and about 100 kg, wears a police cap and reflective vest, and stands on wheeled legs in the middle of the crossing.
Its equipment list, per the project's technical lead at Zhejiang Supcon Information, is lidar plus stereo cameras and a touchscreen on the chest with four entries — directions, report, consultation, education — and it can stay on post for more than eight hours per shift. It is linked to the traffic signal controller and performs standard directing gestures for straight-ahead and left-turn phases, answers visitor questions by voice, and flags non-motor vehicles crossing the stop line, riders without helmets and pedestrians entering on red, delivering a spoken advisory. Police at the junction say it covers peak-hour flow beside a school and frees officers for incident handling. An earlier version appeared at the 2025 summit; the town has built four instrumented 'holographic' digital roads since March 2024 at a reported cost above 30 million yuan.
Our takeThis splits into two jobs with different risk. The perception side — helmets, stop lines, red-light entry — is detection on a fixed camera geometry, where such systems are strongest and false positives are cheap to absorb as a spoken reminder. The directing gestures, by contrast, are slaved to the signal controller rather than decided by the robot. Reports list sensors and functions but not autonomy level, handover, or behaviour when a detector misfires in a dense crowd. That gap, not the height, is the engineering.