What to do if your robot does something stupid?
You can take the help of this new tech developed by the researchers from CASIL, MIT and Boston University. All you need to do is think, and tell the robot that it’s committing a mistake.
The story of Baxter
The team demonstrated their mind-controlled feedback system by putting a robot named Baxter (created by Rethink Robotics) to work. It was assigned with a task to sort and put paint cans and wires into separate boxes. A human, wearing a weird cap (an EEG monitor), was to judge if Baxter is doing things correctly.
Taking data from the EEG monitor, MIT’s feedback system uses machine learning algorithm and understands the human’s mental state when Baxter does the sorting wrong.
According to the research paper published by the team, their system looks for “error-related potentials” brain signals that occur naturally when the robot commits a mistake. It’s something like, “Hey mate, you aren’t doing it correctly.”
It takes around 10 to 30 milliseconds for the system to process the human brain signals and send them to the robot which corrects itself almost instantly. All of this happens without saying a word.
The current system might not be to handle task more complicated than deciphering binary choices. It can open new mediums for the humans to connect with robots. Once systems like these become more capable, according to CSAIL director Daniela Rus, they can help us better supervise factory robots, driverless cars, and other future technologies.