Mind control: PhD student Michele Tavella operates a wheelchair that uses “shared control” to navigate. Brain signals are translated into simple commands like “forward” or “left”; the chair then steers itself around any obstacles.
Artificial intelligence improves a wheelchair system that could give paralyzed people greater mobility. A robotic wheelchair combines brain control with artificial intelligence to make it easier for people to maneuver it using only their thoughts. The approach, known as “shared control,” could help paralyzed people gain new mobility by turning crude brain signals into more complicated commands
MIT Technology Review, September 27, 2010, by Duncan Graham-Rowe — The wheelchair, developed by researchers at the Federal Institute of Technology in Lausanne, features software that can take a simple command like “go left” and assess the immediate area to figure out how to follow the command without hitting anything. The software can also understand when the driver wants to navigate to a particular object, like a table.
Several technologies allow patients to control computers, prosthetics, and other devices using signals captured from nerves, muscles, or the brain. Electroencephalography (EEG) has emerged as a promising way for paralyzed patients to control computers or wheelchairs. A user needs to wear a skullcap and undergo training for a few hours a day over about five days. Patients control the chair simply by imagining they are moving a part of the body. Thinking of moving the left hand tells the chair to turn left, for example. Commands can also be triggered by specific mental tasks, such as arithmetic.
But EEG has limited accuracy and can only detect a few different commands. Maintaining these mental exercises when trying to maneuver a wheelchair around a cluttered environment can also be very tiring, says, José del Millán, director of noninvasive brain-machine interfaces at the Federal Institute of Technology, who led the project. “People cannot sustain that level of mental control for long periods of time,” he says. The concentration required also creates noisier signals that can be more difficult for a computer to interpret.
Shared control addresses this problem because patients don’t need to continuously instruct the wheelchair to move forward; they need to think the command only once, and the software takes care of the rest. “The wheelchair can take on the low-level details, so it’s more natural,” says Millán.
The wheelchair is equipped with two webcams to help it detect obstacles and avoid them. If drivers want to approach an object rather than navigate around it, they can give an override command. The chair will then stop just short of the object.
In Millán’s prototype, 16 electrodes monitor the user’s brain activity. So far it hasn’t been tested on any paralyzed patients.
Damien Coyle, a researcher in the Brain-Computer Interfacing and Assistive Technology group at the University of Ulster, says EEG signals can be slow and tricky to work with. Because of this, he says, many researchers are looking at ways to use shared control, and Millán’s project is a good example of it being put into practice. “The more shared control you have, the better the brain-computer interface, and the faster the person can get from one place to another,” Coyle says.
Millán’s team is developing object recognition capabilities to make the chair smart enough to automatically “dock” with a table or desk to ensure the chair is close enough and not skewed at an angle.