Why scientists want to make robots build Ikea furniture
Disappointment and the anxiety of trying and failing to assemble Ikea furniture may seem like a humiliating exercise to you, but be aware of this: The nightmare of particles can one day lead to robots that are not so stupid.
In recent years, robotics have discovered that building Ikea furniture is actually a great way to teach robots how to handle real-world chaos. A group of researchers coded a simulator in which virtual robot weapons used trial and error put the chairs together. Others managed to get another set of robotic weapons to build Ikea chairs in the real world, however it took him 20 minutes. And now, a handy robot can help a man assemble an Ikea bookshelf by predicting what piece he will want next and deliver it.
“Oneshte is one of those things that is easy to try – even if we break down a few libraries in the lab, it’s not a big deal,” says Stefanos Nikolaidis, a University of Southern California robotist who co-authored a last letter describing the research, which was presented in May at the International Conference on Robotics and Automation. “It simply came to our notice then. And it’s also something we all need to do at some point in our lives. ”
Nikolaidis and his colleagues began by studying how different people build an Ikea bookstore. Instead of providing them with that pictorial instruction sheet, they had the subjects improvise the order in which they configured the support boards for the frame as well as the shelf inserts. (This is an important difference, because the biggest research question for this experiment is not about building furniture – more about it in a second.) Based on these results, researchers can group people into types, or preferences. . Some would merge all the shelves into one of the frames, for example. Others would tie a single shelf to both frames at the same time. These are known as action sequences.
They then had the subjects do the assembly again, this time with a robot arm close by to catch pieces for them. The researcher records which parts (shelves or supports) the person started with, creating a model to see the robot. “Let ‘s say you go in and put the first shelf,” says Nikolaidis. “Okay, the robot doesn’t know that much. Then you select the second shelf. And now start placing the third shelf. Well, you are very, very likely to be in that group of users who have assembled all six shelves in a row. Summer very, very much unlikely since then you would suddenly change your preference. “Once the robot recognizes a person’s preference, it will give them the part they know people like them had previously chosen. Experiments have shown that the robot can quickly and accurately adapt to a human style in this way, distributing successfully the right components.
Think of this as how AI researchers develop an image recognition algorithm: If you want to detect cats, you feed a neural network oodles of feline images. Because he has seen so many examples before, the algorithm can be generalized. If you show her a picture of a cat that has never been seen before, she can rely on her previous knowledge to confirm that she is indeed analyzing a four-legged furry mammal with a crazy attitude.
This robot is doing the same thing, only instead of using a bank with static images, it is relying on examples of sequences, the order in which people joined shelves and backers, based on their preferences. “The robot knows that the next action it has to do is hand you over to the next shelf, certainly very, very high,” says Nikolaidis.
Ultimately, however, this research is not about developing highly specialized robots that come to your home and help you build bookstores. Nor is it about developing machines that can do complex tasks like this on their own. It is about teaching robots how to interact with humans without driving them away more crazy than people already get when building Ikea furniture.
Despite all the hoops about robots coming to steal our jobs, the reality is that you are more likely to do have a working car with you than completely replace. At the moment – and probably for quite some time in the future – people will be much better at certain tasks. No machine can replicate the dexterity of the human hand or come close to solving problems like we do. What robots They are good at is wild work Think of an automotive assembly line: Robotic arms grip car doors in place, but detailed work requires a human touch.