An AI assistant dramatically improves the usability of bionic hands, boosting success rates in delicate tasks and reducing the cognitive load on users. By Jacek Krywko.
The article describes a novel approach to bionic hand control – an AI-powered co-pilot system. Unlike traditional methods relying solely on user input interpreted via EMG signals, this system uses AI to predict the user’s intended actions and assist in their execution. The core methodology involves training a machine learning model on EMG data collected during attempted object manipulations. This model then anticipates the user’s movements, providing subtle corrections and adjustments to the hand’s actuators.
Lab testing with both amputee and intact-limb participants showed a remarkable increase in success rates for delicate tasks. The AI also demonstrably reduced the cognitive effort required to operate the prosthetic, freeing up mental resources for other tasks. Researchers emphasize that while robotics themselves are reaching a high level of dexterity, the bottleneck remains the interface between the user’s nervous system and the prosthetic device.
Challenges include the inherent noisiness of surface EMG and the need for more invasive, yet accurate, neural interfaces. The team is actively pursuing research into internal EMG and neural implants to improve signal quality and control precision. They also seek industry partnerships to move the technology from the lab to real-world clinical trials and eventual commercialization. Nice one!
[Read More]