The robot can perform tasks through the brain waves of the operator and thinking commands

NetEase Technology News, June 21, according to VentureBeat reports, the use of thinking to control robots is not as remote as it sounds. Researchers at the Computer Science and Artificial Intelligence Laboratory (CSAIL) at the Massachusetts Institute of Technology (MIT) have developed a new interface that can read the brain waves of human operators and let them order machines to perform tasks by thinking.

Figure: The robot system of the Computer Science and Artificial Intelligence Laboratory (CSAIL) under the Massachusetts Institute of Technology (MIT) monitors the brain waves of human operators

Daniela Rus, director of CSAIL, said: "We want to stay away from a world where people have to adapt to machine constraints. Methods like this show that it is very possible to develop a more natural and intuitive robotic system." The system uses a combination of electroencephalography (EEG) and electromyography (EMG) to monitor brain activity. The electroencephalogram detects electrical activity in the brain through electrodes connected to the scalp, and the electromyogram is used to measure Signals produced by motor neurons.

Neither EEG nor EMG is perfect science, nor is it particularly accurate. However, by combining the two, the research team was able to obtain higher accuracy than using one of the techniques alone. Joseph DelPreto, a doctoral student and the first author of the project's thesis, said: "By looking at muscle and brain signals, we can start to understand a person's natural posture and what they have done to determine whether something went wrong Hasty decisions. This helps people to communicate with robots to become more like people to people. "

The CSAIL team's algorithm analyzes the signals of "error-related potentials" (ErrPs), a pattern of neural activity that occurs naturally when people notice errors. At the moment when an error is detected, such as when the controlled robot is about to make an error, it will stop working so that the operator can use the gesture-based menu interface to correct the error.

Ruth said: "This work that combines EEG and EMG feedback allows the interaction between humans and robots to be applied to a wider range of applications, which is more than we can only use EEG feedback before. More needs to be done. By including muscle feedback, we can use gestures to remotely command the robot, resulting in more nuances and specificities. "

The researchers found that compared to 70% of the control group, human-supervised robots took more than 97% to correct errors. Even more impressive is that this system is equally effective for people who have never used it before. The research team believes that the system may be more useful for people with language impairments or limited mobility.