Simple Neural Networks Outperform Complex Systems in Controlling Prosthetic Limbs

Simple neural networks are more effective than complex systems in controlling robotic prostheses
Researchers at the University of Michigan found that artificial neural networks based on the natural nerve circuits of the human body gave primates faster and more accurate control of prosthetic fingers and hands controlled by the brain. Researchers at the University of Michigan have shown that artificial neural networks inspired by natural nerve circuits in the human body give primates faster and more accurate control of brain-controlled prosthetic hands and fingers.

A team of doctors and engineers found that using a feed-forward network to control robotic fingers improved the peak finger velocity of the robot by 45% compared with traditional algorithms without neural networks. The team of engineers and doctors found that a feed-forward neural network improved peak finger velocity by 45% when compared to traditional algorithms not using neural networks.

This feed-forward architecture is older and simpler, with information flowing only one way, from input into output, said Cindy Chestek Ph.D. an associate professor in biomedical engineering, U-M, and the corresponding author for Nature Communications.