Successfully developed robots that can be controlled by thought
Robots can teach themselves new tips through machine learning. But now researchers at the Computer Science and Artificial Intelligence Labs, Masachusette Technology College (MIT) and Boston University (BU) add another way to manipulate robots. Use your brain.
Researchers have created a feedback system that can read human brain responses when looking at an active robot. If the robot goes wrong, the operator will recognize and the robot is sent a signal to correct the error. The researchers shared the results of the study in a report published today.
Robots can teach themselves new tips through machine learning.
No one button or say a word
In order to communicate electronically with a machine, researchers have developed an EEG cap for the operator to record brain activity. Then, in real time, the robot receives the nerve signals from the operator and can change the way it operates.
The research team, led by MIT laboratory director Daniela Rus and BU professor Frank Guenther, found they could detect when a person thought the robot had done something wrong by following it. The brain signal is referred to as the "error-related signal" (ErrP). ErrP is generated when we recognize a mistake, and then we emit electronic signals and the EEG helmet can collect, sort, and send as a kind of feedback to the machine. Just make a mistake and let it fix it yourself.
If the brain does not send out ErrP signals, the robot will continue to do what it s doing before.
In the video, this telepathic "robot" robot only works in a simple situation with two possible outcomes. But researchers hope someday that refined systems will be able to interact in more complex tasks someday.
This is not the first time researchers have tried to develop brain-driven robots, but previous efforts have required the operator to think in a specific way that the robot can understand. .
"As you can see in the robot, all you have to do is agree or disagree with what it is doing," Rus said in a statement. "You do not have to train the brain to think in a specific way, the machines will follow your thinking, not the opposite."