Have you ever seen a robot blush? Baxter, the chunky red robot with a face, can do that … when he makes a mistake. But that’s OK – he’s learning. He’s at school. His teachers and trainers are a combined group from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Boston University.
Baxter is the product of Rethink Robotics. The company advertises “smart, collaborative robots” for use in manufacturing and industry. This two-armed model was preceded by Sawyer, a one-armed version with strong sensing capabilities allowing it to work safely next to humans in collaborative settings.
They’re using an electroencephalography (EEG) cap to read a human observer’s signals in real time. The brain signals help Baxter correct his errors as he performs simple sorting tasks. The human controller sports a flexible cap covered with red nodes and wires. It looks more akin to something you might find at the beauty salon than in most electronics labs. However, it’s been used for years as a non-invasive way of collecting brain signals and converting them to digital form.
It’s all part of discovering innovative ways for humans and machines to communicate. In the past, non-invasive EEG reading has been inaccurate. Special training has been required to use the equipment. New variations using signals occurring when the wearer observes an error are now being tested.
The signals, called Error Related Potentials, or ErrP for short, can be easily recognized by their distinctive shape. Because of this, scientists are looking at ways humans can use them to increase accuracy when teaching machines. This new process is so much more effective, that more people will able to use it without special training.
A video from the MIT lab shows the human controller sitting opposite Baxter. The robot’s rectangular face looms above the table as his tubular arm moves from side to side. A hand with four white grippy fingers reaches for wire rolls and spray paint cans. It places each object in one of two labeled bins. Communication is close to instant. The machine classifies brainwaves in as little as 10 milliseconds. When it makes a mistake, 2 circular pink splotches bud on its flat screen “cheeks.” It then corrects the task.
Now the question is … what can a robot like this do for you?