Researchers have developed an impressive full-body, wearable robotic ‘exoskeleton’ with 208 muscles.
Exoskeletons are wearable robotic frameworks for the human body.
They promise easier movement and improved mobility for people with disabilities, as well as safer and more efficient movement for factory workers and astronauts.
To date they have had to be pre-programmed for individuals and specific activities based on lengthy and costly tests with human subjects.
But now, the scientific journal Nature describes a super-smart, ‘learned’ controller that leverages data-intensive AI and computer simulations to train them, as reported by What’s The Jam.
“This new controller provides smooth, continuous torque assistance for walking, running or climbing stairs,” Dr. Shuzhen Luo of Embry-Riddle Aeronautical University, Florida, US, said.
“With only one run on a graphics processing unit, we can train a simulation, so that the controller can effectively assist all three activities and various individuals.”
Driven by interconnected, multi-layered neural networks, the controller learns as it goes.
Dr. Luo said: “It evolves through millions of epochs of musculoskeletal simulation to improve human mobility.”
Used on a custom hip exoskeleton, this learning framework has generated an average fall in energy expenditure for the wearer of 24.3% for walking, 13.1% for running and 15.4% for climbing stairs – the highest reduction to date.
The reduction rates were calculated by comparing the performance of a human with and without the robotic exoskeleton assistance.
Dr. Hao Su of North Carolina State University, US, said: “It’s a true measure of how much energy the exoskeleton is saving.
“This work is essentially making science fiction a reality — allowing people to burn less energy while conducting a variety of tasks.”
The approach is believed to be the first to demonstrate the feasibility of developing controllers that bridge the simulation-to-reality gap while also significantly improving human performance.
Dr. Luo added: “Previous achievements in reinforcement learning have tended to focus primarily on simulation and board games.
“We proposed a new method – a dynamic-aware, data-driven reinforcement learning way – to train and control wearable robots to directly benefit humans.”
It is hoped the framework will offer a “general” and “scalable” strategy for the deployment of assistive robots to both able-bodied and mobility-impaired humans.
READ MORE: Newly-wed couple opt for kebab-cutting ceremony instead of cake