The researchers' new approach uses AI and computer simulations to train a robotic exoskeleton to help its user conserve energy autonomously.
In a new study, researchers from North Carolina State University show that this technologically advanced device is the result of reinforcement learning, a technique for training software to make decisions.
In a demonstration video provided as part of the new study published in the journal Nature, the method consists of integrating three neural networks: a movement imitation, a muscle coordination, and an exoskeleton control network.
Subjects began walking, running, and climbing stairs, demonstrating smooth transitions between the three sets of movements. It took just eight hours for the machine to adapt to the body.
While you run, the device intelligently senses the support your body needs.
With the added benefit of reducing energy expenditure while walking by 24.3%, while running by 13.1% and while climbing stairs by 15.4%, the device may be of interest to able-bodied individuals looking to improve their performance and training.
“This research essentially makes science fiction a reality, allowing people to reduce the amount of energy they use to perform a variety of tasks,” Su said.
A leap in machine learning technology
The device is targeted at both able-bodied and movement-impaired people and does not require laboratory or clinical environments, likely because it is a highly intelligent technology that relies on the individual wearing it.
“Previous work has tended to focus primarily on simulations and board games,” Shuzhen Luo, lead author of the paper, told North Carolina State University.
But this solution, combining AI and computer simulation, lays the foundation for the future of wearable robots — in other words, they will become smarter and more usable soon.
A new frontier in machine learning has been opened.
Researchers at North Carolina State University say exoskeletons that look like something out of a sci-fi movie are in the early stages of development for older adults, people with neurological conditions like cerebral palsy or amputees.
Hopefully, it will be available soon for those in need of better support. While the actual design may require adjustments based on their needs, the potential of this machine learning technology is very appealing not only for injured individuals but also for able-bodied individuals looking to optimize their health.
“Exoskeleton assistance without experimentation through simulation learning” has been published in the journal Nature.
About the Editor
Maria Mocerino A Los Angeles native, Maria Mocherino has had articles published in Business Insider, The Irish Examiner, The Rogue Mag, Chacruna Institute for Psychedelic Plant Medicines, and now Interesting Engineering.
