DEEP Robotics tests its DR02 humanoids as traffic guards at rainy Hangzhou intersections

A staff member at the Hangzhou trials, in a quote DEEP Robotics supplied, said "with many repetitive advisory tasks handed to the robot, we can devote more energy to accident handling and enforcement of key violations," and called that "work that requires more human judgment." ^[9]

The staff member is describing a hand-off of chores. DEEP Robotics frames the robot the same way: it takes routine inquiries while human staff keep emergencies, enforcement and the calls that need judgment ^[8].

The thing being pitched is a humanoid at an intersection. The thing being done, on that staff member's account, is repetitive advisory work ^[9]. TechRepublic, which reported the trials, wrote that a humanlike form can use existing pedestrian space and hand signals without anyone redesigning the street, but that this does not make humanoids the most effective or affordable option ^[12]. Fixed cameras, automated warning systems and conventional traffic-control equipment already handle some of the same functions, the publication wrote ^[11].

The engineering claim the trial tested most directly is weatherproofing. The DR02 carries an IP66 rating against dust and water, which DEEP Robotics tested under sustained rainfall ^[3], in a district where heavy tourist traffic was also straining traffic management ^[1]. Whether the robot did its tasks accurately is a separate matter. DEEP Robotics did not provide trial results on accuracy, uptime, cost savings or traffic flow ^[10].

For a city deciding whether to extend a trial like this, I'd split the robot into its tasks and put each on two axes: whether existing equipment already does it, and how much a wrong output costs.

Multilingual help for tourists sits in the best quadrant ^[7]. A fixed camera does not answer a visitor's question about transport or traffic restrictions. A wrong answer costs that visitor some time. If the humanoid has a case in these trials, it is here.

Violation reminders sit in a harder one. The four violations DEEP Robotics says the robot can recognise ^[6] are, in my view, where its work overlaps most with the cameras and warning systems TechRepublic named. The publication also listed recognition accuracy in heavy rain and the consequences of incorrect warnings among the open questions ^[14].

Directing moving vehicles carries the highest cost of a mistake, and conventional traffic-control equipment already covers part of that job ^[11], so I would put it last in any extension.

Each task a city keeps comes with two numbers to ask the vendor for: the error rate in rain and the hours the robot actually worked. TechRepublic wrote that any broader rollout depends on measurable benefits over existing traffic-management systems ^[16]. For now, it wrote, the value to city authorities is testing whether robots can supplement existing staff during peak periods and difficult weather ^[17].