Xiaomi has made its car factory cause a sensation. With the SU7 and YU7 on the market, the company has managed to get people to line up and pay a fortune to see how these cars are manufactured, something they have achieved thanks to a good marketing strategy and, above all, to a futuristic factory with a hook that is difficult to ignore: humanoid robots riding the cars.
It was this past March when we learned from Lei Jun, founder and CEO of the company, that they had put a series of humanoid robots to “do practices” inside the factories. The tasks of these robots were not at all exciting, as they were dedicated to one thing: putting self-tapping nuts on the floor parts of cars.
However, the period of being robo-interns is coming to an end and it was the CEO himself who commented that robots have climbed several steps. Specifically, they are almost on the same level of success as their human “companions.”
From robo-interns to robo-experts
The task they had to do is the most alienating and example of chain work that you can imagine. The system collects the nuts from an automatic supply equipment and deposits them in the indicated place, where the automatic screwing is subsequently carried out. Indeed: the robots were putting nuts… and that’s it.
In March, Lei Jun said the robots had achieved a 90.2% success rate in placing nuts during three hours of continuous autonomous operation.
This figure indicates the number of correct operations compared to the total attempts made and, in addition, they detailed that they adjusted to a production cycle of 76 seconds. Timing in these factories is crucial, especially when all steps are so automated if the car output goals are to be achieved.
Now, although putting on nuts seems simple, they are not like those on an Ikea piece of furniture and there are some things to consider. A key challenge in this process is getting the self-tapping nut to precisely align, center and seat on the locating pin before screwing. Otherwise, the system crashes and the join is incomplete.
And this process is complicated by three main factors:
- Grip variability: The orientation of the nut on the robot gripper changes with each cycle.
- Design and physical forces– The internal knurling of the nut and the magnetic attraction of the pin make exact adjustment difficult.
- Operating environment: There are limitations on the working angle and external interference in the assembly area.
For the robot it is more complicated than it seems, but now it has been Lei Jun himself who has come to the fore again to point out that robots have gone from that success rate of 90.2% to 98%. They are, according to the executive, just one percentage point behind human workers. And most importantly, they are ready to take on more complex tasks.
Because thanks to that success, it has been decided that Robots can begin to tackle more complex tasks. Because from tightening nuts, they have moved on to new stations that included tasks such as installing the side panels of the center console, as well as folding and recycling parts. The side panel task is complex because it is a large, flexible, irregularly shaped component that requires multiple pick-and-place cycles in different positions.
Xiaomi says this is the first time a humanoid robot has completed prolonged manipulation of such a part on a real automotive production line and estimates the success rate to be over 90%. In four months we will see where that percentage is, but although it is clear that the evolution of the company’s humanoid tobots It is being remarkable, we are talking about robots that are still in the testing phase and in a single workstation.
They are taking data and running every possible metric to determine if they are up to par for a productive, full-scale deployment. Human workers can rest easy, at least for a few more months.
What I don’t understand is why they have to look like robots taken from the reboot by ‘Robocop’.
In | In the war of humanoid robots, those from the United States dance and those from China work by the piece. It is not a technological issue
