Shulou(Shulou.com)11/24 Report--

Flying Spaghetti Monster, is that you?

In the sky, there is a monster that we can't see flying.

The main ingredient of its body is starch, specifically a loose piece of spaghetti. Among the messy noodles, two eye stalks like snails are protruding, holding up a pair of big eyes that are absent-minded but irresistible. In addition, the soft-looking noodles hold on to two full meatballs, a common spaghetti side dish.

The monster, called Flying Spaghetti Monster, was created by a recent American college graduate, Bobby Bobby Henderson, in 2005.

Bobby, who studied physics at Oregon State University, came up with the Flying Spaghetti Monster to express protest and ridicule when he learned that the Kansas Board of Education had decided to allow public high schools to offer courses related to creationism. On this basis, he founded the Flying Noodle Religion (Pastafarianism), the doctrine that the universe was created by this invisible monster-and Bobby called in an open letter to the Board of Education that school science classes should pay enough attention to the creator.

Maybe it's just a coincidence. Recently, a robot appeared, its body looks like a lot of noodles gathered together, and it will also use those noodle-like tentacles to hold on to a ball.

Image source: how the original paper "Noodles" is rolled up in the 21st century robot, can run, can dance, can go deep into the ruins of nuclear power plants to detect radiation levels, can explore unknown space on the moon. It seems that robots can do what human beings can do, and robots can do what human beings can't do. However, before the total rout, humans still have some unique skills that have not yet been conquered by robots.

One of them is crawling skills. Whether it's a mug or a paper cup, a football or a balloon, a banana or a bunch of grapes, we can easily pick up all kinds of objects. according to its shape and texture to decide what kind of force to apply-this is not usually a complex calculation for humans, and the brain may have made the decision long before we know it.

Photo: Tenor, but for robots, grabbing objects is never a simple and straightforward task. Especially when it comes to irregularly shaped objects, they rarely make proper grabbing movements, and they are often clumsy and funny, reminiscent of some doll machines that are deliberately set hell parameters and can't catch the doll.

Over the years, there have always been many scientists challenging this difficult problem in the world. As a result, we will see in the science and technology news every now and then that new robots are born for crawling, and their skills have reached a higher level. In October, a noodle-shaped tentacle robot appeared in front of people.

When inflated, the internal pressure of 172kpa allows "noodles" to be highly curled (photo: original paper). It comes from the Harvard School of Engineering and Applied Sciences (SEAS) and can be entangled by a set of slender tentacles. Scientists say it is similar to jellyfish, which stick their tentacles to their prey. For robots, one tentacle is weak, but with 12 tentacles working together, it is easy to hold on to objects and irregularly shaped "prey" can also cope.

However, after the robot's tentacles catch the prey, they do not release toxins that paralyze the prey as many jellyfish do. After all, it is not designed to hunt any kind of animal, grasping and carrying are the main functions. When there is no work, the robot's tentacles will hang down naturally, but when the work comes, the tentacles can quickly curl up and grasp the object.

These smart tentacles are controlled by inflation and deflation. By implication, they are hollow, more like Bucatini in the spaghetti family. Without inflation, these noodles are in a relaxed state, and when inflated, they will curl and become grabbing tools.

The wall of the tube is thick on one side and thin on the other. (photo: original paper, Harvard Microrobotics Lab / Harvard SEAS) each "straw surface" is about 30 cm long and is made of silicone rubber and is an elastomer. The key point is that the thickness of the pipe wall is not uniform, one side is thicker and the other side is thinner. In this way, the "straw surface" can become highly bent after being inflated, so that more areas can come into contact with the object it is trying to grab. The 12 noodles are arranged in two circles, the outer circle is 5 cm in diameter and the inner ring is 2.5 cm in diameter, which together make up the robot's grip.

When the robot grabs the object and carries it to its destination, and the operator wants it to let go, as long as it deflates the noodles, it will return to its natural drooping form.

When grasping different objects, the force required is always different. It can be said that each object brings a new task, which may require specific analysis of specific problems. In many previous studies, scientists often use complex algorithms to determine what action to use for each grab according to the properties of the object or the state of the robot when it comes into contact with the object.

However, this noodle robot does not need intelligent action planning, does not rely on perceptual or feedback data to adjust posture, eliminates the process of machine learning, and can be inflated within a certain range of sizes. grab objects of all shapes.

What can it catch if it doesn't change?

Let's start with objects with relatively simple shapes. The grab shown at the beginning is just one of them, and the scientists let the robot dominate a hollow cylinder. When the cylinder lies flat, the rolled-up robot can hold it firmly and transmit it to the designated position; when the cylinder stands up, it is also easy for the robot to grab it from the inside.

Inflating, grabbing the cylinder, transferring to a specific location, and letting go (photo source: original paper) another grasping object of the robot is a small tree-like shelf. It has eight branches, is more complex than the shape of a sphere or cylinder, and has more irregular boundaries for robots to touch-which, in the eyes of researchers, represents a new grasping task that is essentially different from traditional grasping.

But in front of the robot's grip, the new task doesn't seem to pose an extra difficulty: the success rate of 80 grab tests is more than 90%. As a result, the scientists set an additional question, not allowing the robot to grab directly from the top of the tree, but shifting a distance to the side (such as 1cm, 2cm, 3cm) and then entangling the target.

The results show that the influence of slight dislocation is not great, and the noodle robot has a high tolerance for position error, when the distance between the central axis of the grip and the central axis of the target object is as far as 0.6 radius of the object, the success rate of grasping (and short distance handling) is still high. If you want to shake the robot's performance, you need to shift the grip to a farther position before you start to act.

Scientists are gratified by the excellent ability of the noodle robot to grab from the top on the left, to offset 0.6 radius of the object in the middle, and to grab the noodle robot after deviating from the radius of 1 object but failed (photo source: original paper). If you think that the "tree" in the laboratory is nothing, there will be more complex goods waiting for it to dominate in the real world.

For example, many people like to raise flowers and plants at home, and small potted plants have become the goal of robots. When the internal pressure is 172 KPA, the curly noodles smoothly pick up a plant.

Photo source: there are several other potted plants in the original paper, although each has its own form, but none of them can escape a similar fate. Some irregularly shaped toys, as well as a small tripod, were also captured by robots in the same way. In response to changes, what the researchers appreciate most is the adaptability of this robot. It may not create everything like the legendary spaghetti monster, but it can play with many different objects with 12 flexible tentacles.

Photo Source: the birth of this robot was not achieved overnight. Before the mold manufacturing, the researchers have tested the ability of the noodle robot with simulation experiments on the computer: to see how the noodle grips should be arranged, or how big the two circles should be, in order to achieve a higher success rate of grasping.

The simulation experiment (photo source: original paper) it probably failed countless times in the virtual world before it succeeded in catching prey in the real world. Of course, there will still be many failures in the real world, and scientists still need to continue to work hard.

If you encounter a verification program that proves that you are not a robot while surfing the Internet, telling it that you are better at grasping objects than a robot will not help you pass the verification.

Original paper:

Https://www.pnas.org/doi/10.1073/pnas.2209819119

Reference link:

Https://www.seas.harvard.edu/news/2022/10/tentacle-robot-can-gently-grasp-fragile-objects

Https://www.nytimes.com/2005/08/29/arts/design/but-is-there-intelligent-spaghetti-out-there.html

Https://www.nsri.org.za/2012/02/how-jellyfish-sting/

This article comes from the official account of Wechat: global Science (ID:huanqiukexue), written by: chestnut, revision: clefable

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.