Parasol Island – Integrated Digital Studio

AI-assisted robotic prostheses enable paralysed patients to walk.

A team of researchers in Canada have achieved a milestone in the development of mechanical prostheses. The ExoNet project has worked to create an intelligent leg prosthesis which can recognise its environment and perform appropriate actions without human instruction. “Similar to autonomous cars that drive themselves, we’re designing autonomous exoskeletons and prosthetic legs that walk for themselves”, says Brokoslaw Laschowski, project manager and a Ph.D. candidate at the University of Waterloo in Ontario.

Motorised prostheses are not new to the market, but until now they had to be manually operated via smartphone apps, joysticks or electrodes on the body. The ExoNet prosthesis, equipped with cameras and sensors, combines computer vision with Deep Learning for the first time in order to identify desired movement sequences. The system draws its knowledge from a specially created open-source database with more than 5.6 million high-resolution RGB images of real-world indoor and outdoor environments which were recorded first-person while walking. “We are providing the robotic exoskeletons with vision so that they can navigate for themselves”, explains Laschowski. The person wearing the prosthesis no longer has to trigger actions themselves such as walking, standing, sitting or climbing stairs.

The AI prostheses can be used by people with multiple sclerosis, spinal cord injuries and other ailments which prevent them from walking independently. ExoNet allows forward motion at one metre per second – the same speed as a person without a disability.