The more the merrier?
Editor’s note: This article is from the WeChat public account “Big Data Digest” (ID: BigDataDigest), source IEEE, compile Zhao Wei.
Speaking of group robots, there is a huge robot army in the minds of Digestococcus. Each robot is independent and smart. However, the group robots we mentioned today are stupid. The small robots are made up of one another, and they don’t have any skills when they are independent individuals. They can also perform complex tasks in combination.
The original intention of group robots is to replace a single, expensive, fragile single-task robot with a simpler, cheaper (marked) and replaceable robot that can work together to accomplish the same type of task. . But the problem of building such a group robot is that if you want them to accomplish a specific task, they need to calculate and communicate with each other.
Another implementation of group robots is to use a cheaper but less intelligent group of robots. In fact, if you use some of their physical features to drive them, they may not need to be intelligent at all. These group robots are “random,” meaning that their actions are randomly determined, but if carefully designed, they can still be tasked with specific tasks.
Unrelated small robots can work together
Not long ago, Georgia Tech developed a small robot called “smaticles”: they are simple in design and single in function, but once combined, their randomness can be given to the entire group. The ability to complete certain tasks.
To be honest, the “smart” of these particle robots may be a bit exaggerated. In fact, they are “a bit stupid”. Strictly speaking, their ability to complete tasks solely by their own power is almost zero. These smart particles weigh a single 35 grams and consist of some small pieces of 3D printed on the server, an Arduino Pro Mini, a battery and a light or sound sensor. When small fragments are activated, each smart particle can move slightly; but a single intelligent particle has a large probability of moving only in a square and then gradually drifts over time.
When a group of intelligent particles are in a restricted area, the situation changes: a small number (five or ten) of intelligent particles are grouped together to form a “super-smart particle.” In addition to being very close to each other, smart particles do not communicate with each other, as far as each smart particle is concerned.They are independent, but the magic is that some of them work together without any connection. “These are very simplistic robots whose behavior is dominated by the laws of mechanics and physics,” said Dan Goldman, professor of physics at Georgia Tech.
A flashlight can guide the “robot” movement
The researchers noticed that if one of the small robots stopped moving for reasons such as exhaustion, the intelligent particle swarm would start moving in the direction of the stationary robot. Graduate student Ross Warkentin learned that he can control the movement by adding a light sensor to the robot: when a strong beam hits one of the robots, it stops swinging its arm.
“If you adjust the flashlight to the right angle, you can highlight the robot you want to be at rest and move it toward or away from the light, even if the robot is not pre-programmed toward the light,” Goldman said. “This way you can guide the whole movement in a very basic and random way.”
It turns out that people can model this behavior and control super-smart particles through the maze with enough fidelity. Strictly speaking, these special smart particles are not particularly small, but the original intention of the researchers was to develop related technologies. When the robots were scaled down to the extent that we could not plug them into computing devices, these intelligent control technologies would still kick in.
Fancy combination of small robots?
The researchers are still working on other concepts, such as:
smarticle: smart particles; suoersmarticle: super-smart particles; slug: peristaltic super-robot group; biped: biped-like super-robot group
In the above picture, Georgia Tech’s researchers envisioned a group of random robots that did not have well-defined shapes or contours but were able to take advantage of overall level behavior and collective movement to achieve self-propulsion. The researchers explained that, as observed in the bio-collectives, complex targets could be achieved in such a group robot.
Hey, I’m not sure if I really want a bipedal robot made up of a bunch of small robots. This looks a bit strange… what do you think?
Related reports:
https://specTrum.ieee.org/automaton/robotics/robotics-hardware/georgia-tech-smarticles-stochastic-robot-swarm