An innovative AI and simulation-based tutoring system for robotic exoskeletons has significantly cut energy usage during tasks like walking, sprinting, and ascending stairs.
North Carolina State University investigators have crafted a novel method that utilizes artificial intelligence (AI) and virtual simulations to endow robotic exoskeletons with the ability to autonomously aid users in conserving energy during movements such as walking, sprinting, and stair climbing.
“Our initiative presents and validates a new artificial intelligence algorithm that narrows the chasm between virtual simulations and the real world, allowing for the autonomous governance of wearable robots for the enhancement of human mobility and health,” remarks Hao Su, the principal author of the study which is set to be released today (June 12) in the academic journal Nature.
Advancing Human Movement Performance
“Robotic exoskeletons hold great promise for augmenting human locomotive capabilities,” states Su, who is an associate professor in the domain of mechanical and aerospace engineering at North Carolina State University. “Nonetheless, their progress and widespread distribution are hindered by extensive human trials and the crafting of control algorithms.
“The fundamental concept is that the AI embedded within a mobile exoskeleton learns to assist individuals with walking, sprinting, or ascending in a virtual environment, eliminating the need for experimental trials,” Su explains.
Particularly, the researchers placed emphasis on advancing the independent governance of embodied AI constructs – which merge an AI module within a tangible robotic framework. This research was principally concerned with coaching robotic exoskeletons to support individuals with assorted bodily movements. Ordinarily, users must undergo extensive hours of “training” with an exoskeleton for the device to discern the necessary force magnitude and the precise moment for force application to aid in ambulation, sprinting, or staircase climbing. Their methodology permits the immediate employment of the exoskeletons by users.
Enhancement of Energy Efficiency via Robotics
“This endeavor is essentially turning what was once science fiction into tangible reality – it permits individuals to exert less energy while performing diverse actions,” Su asserts.
“We’ve engineered a method that conditions and manages wearable robots for the direct advantage of humans,” mentions Shuzhen Luo, the lead author of the paper and a previous postdoctoral fellow at NC State, who now holds the position of assistant professor at Embry-Riddle Aeronautical University.
As an instance, during tests with human subjects, the team uncovered that the participants reduced their metabolic energy usage by 24.3% while walking with the robotic exoskeleton compared to sans it. The energy conservation was 13.1% during a run, and it reached 15.4% during stair climbing.
“It is critical to recognize that these reductions in energy are gleaned from comparing the efficacy of the robotic exoskeleton to an individual devoid of an exoskeleton,” Su clarifies. “This indicates it is a genuine reflection of the energy the exoskeleton conserves.”
While this investigation was primarily with fit individuals, the innovative method is also pertinent to robotic exoskeleton utilizations aimed at facilitating those with mobility challenges.
Potential Directions and Future Evaluations
“Our schema may represent a universally applicable and extensible approach for the swift conception and broad implementation of various assistive robotic technologies, catering to both the able-bodied and those with mobility constraints,” Su mentions.
“Currently, we are at initial stages of validating our novel method’s effectiveness in robotic exoskeletons deployed for the elderly and those with neurological disorders like cerebral palsy. Further, we are keen to investigate how this approach could enhance robotic prosthetic device efficiency for individuals with limb amputations,” he continues.
Reference: “Experiment-free Exoskeleton Assistance Via Learning in Simulation” 12 June 2024, Nature.
DOI: 10.1038/s41586-024-07382-4
The study includes contributions from Menghan Jiang, Junxi Zhu, and Israel Dominguez Silva, doctoral candidates at NC State; Sainan Zhang and Shuangyue Yu, postdoctoral associates at NC State; Tian Wang, a graduate scholar at NC State; Elliott Rouse from the University of Michigan; Bolei Zhou from the University of California, Los Angeles; Hyunwoo Yuk from the Korea Advanced Institute of Science and Technology; and Xianlian Zhou from the New Jersey Institute of Technology.
This exploration was supported by the National Science Foundation under grants 1944655 and 2026622; the National Institute on Disability, Independent Living, and Rehabilitation Research, under grant 90DPGE0019 and Switzer Research Fellowship SFGE22000372; and the National Institutes of Health, under grant 1R01EB035404.
Shuzhen Luo and Hao Su are inventors on a patent pertaining to the control mechanism discussed in this study. Furthermore, Su is a co-founder and holds financial interest in Picasso Intelligence, LLC, a company that develops exoskeletons.
Image Source: PeopleImages.com – Yuri A / Shutterstock
