How do you tell your robot not to do something that could be catastrophic? You could give him a verbal or programmatic command or you could watch him watch your brain for signs of distress and have him stop himself. That's what the researchers at MIT's robotics research laboratory did with a system that is connected to your brain and tells robots how to do their job.
The initial system is quite simple. An EEG and EMG scalp system is connected to a Baxter work robot and leaves a wave or human gesture when the robot does something that it should not do. For example, the robot can regularly do a task – drill holes, for example – but when it approaches an unknown scenario, the human can make a move to the task that should be done.
"By examining the muscular and cerebral signals, we can begin to become aware of a person's natural gestures and hasty decisions about whether something is wrong," said doctoral candidate Joseph DelPreto. . "It helps to communicate with a robot rather than communicate with another person."
Because the system uses nuances such as gestures and emotional responses, you can train robots to interact with people with disabilities and even prevent accidents by taking concerns or alarms before communicating them verbally. This allows workers to stop a robot before it damages anything and even help the robot understand slight changes in its tasks before it starts.
In their tests, the team trained Baxter to drill holes in an aircraft fuselage. The task changed from time to time and a human standing nearby was able to signal to the robot to change position before drilling it, essentially causing him to do new tasks in the middle of his current task. In addition, there was no real programming on the part of the human, just a suggestion that the robot moves the drill left or right on the fuselage. The most important thing? Humans do not have to think in a special way or to train to interact with the machine.
"What's great about this approach is that it's not necessary to train users to think in the prescribed manner," DelPreto said. "The machine fits you, not the opposite."
The team will present its findings at the Robotics: Science and Systems (RSS) conference.