Researchers Train Monkeys to Control Motorized Wheelchairs Using Only Their Thoughts

Researchers at Duke University have developed a new brain-machine interface that allows a pair of monkeys to control motorized wheelchairs with only their thoughts. This development is intended to provide an alternative method of controlling the mobility devices for physically disabled people, as some individuals are unable to blink their eyelids to direct a wheelchair’s controls.

The research team’s brain-machine interface (BMI) made use of hundreds of hair-thin microfilaments that were implanted in the somatosensory and premotor regions of the brains of two rhesus macaques, used to record the signals transmitted by 300 neurons in each of the monkeys. The monkeys were then trained to think about moving toward a bowl of grapes: at first, training was passive, with the researchers pushing them in their wheelchairs toward their goal, while their thought patterns were recorded during the process. The researchers then developed a computer program to control the wheelchairs using the patterns from the macaques’ brain scans, and began training the monkeys to navigate using the chairs on their own.

The monkeys were not only successful in learning how to control the wheelchairs using thought control, but they also provided the researchers with a surprise in the process: their brain signals indicated that they also began contemplating the distance to the food, a development that was unexpected when the experiment was started.

“This was not a signal that was present in the beginning of the training, but something that emerged as an effect of the monkeys becoming proficient in this task,” remarks study co-author Miguel Nicolelis. “This was a surprise. It demonstrates the brain’s enormous flexibility to assimilate a device, in this case a wheelchair, and that device’s spatial relationships to the surrounding world.”

The researchers plan to hone their control method’s accuracy and fidelity in further tests, before applying it in human trials. The team has previously been able to record the signals from up to 2,000 neurons, using the same technique.