summary: A new study details how AI and computer simulations can be used to train a robotic exoskeleton to help users save energy when walking, running or climbing stairs, eliminating the need for lengthy experiments involving humans and could be applied to a range of assistive devices.
This groundbreaking invention has great potential to assist people with mobility impairments and improve the accessibility of everyday life. Researchers found that participants achieved energy savings of up to 24.3% by using the exoskeleton.
Key Facts:
- AI and simulation allow exoskeletons to be trained without human-involved experimentation.
- The exoskeleton allowed users to save up to 24.3% energy in motion tests.
- This method can be applied to a variety of assistive devices, including prosthetic limbs.
sauce: New Jersey Institute of Technology
A team of researchers has demonstrated a new way to use AI and computer simulation to train a robotic exoskeleton to help users conserve energy when walking, running and climbing stairs.
As described in published studies NatureThis new method allows the rapid development of exoskeleton controllers to assist with locomotion, without relying on lengthy experiments involving humans.
Furthermore, this method can be applied to a variety of assistive devices beyond the hip exoskeleton demonstrated in this study.
“This could also be applied to knee or ankle exoskeletons or other multi-joint exoskeletons,” said Xianlian Zhou, associate professor and director of NJIT's BioDynamics Institute.
Moreover, it could be applied to above-knee and below-knee prosthetics as well, bringing immediate benefits to millions of able-bodied and mobility-impaired people, he said.
“Our approach is a major advancement in wearable robotics. Our exoskeleton controller is developed exclusively through AI-driven simulation,” Zhou explains. “What's more, the controller can be seamlessly transferred to hardware without the need for additional human testing, eliminating the need for experimentation.”
This groundbreaking technology has the potential to assist people with mobility challenges, such as the elderly and stroke victims, without the need for extensive laboratory or clinical testing, ultimately paving the way for restoring mobility and improving accessibility in daily life and the community.
“This work proposes and demonstrates a new method using physics-based, data-driven reinforcement learning to control wearable robots that can directly benefit humans,” said Hao Xu, corresponding author of the study and associate professor of mechanical and aerospace engineering at NC State.
Exoskeletons have the potential to improve human athletic performance for a wide range of users, from injury rehabilitation to permanent assistance for people with disabilities, but extensive human testing and regulatory laws limit their widespread adoption.
The researchers focused on improving the autonomous control of embodied AI systems, which are systems in which AI programs are integrated into physical technology.
The research, which focuses on teaching a robotic exoskeleton how to assist able-bodied people in performing various movements, expands on previous reinforcement-learning-based work on lower-limb rehabilitation exoskeletons, also a collaboration between Zhou, Su and several others.
“Previous work in reinforcement learning has tended to focus primarily on simulations and board games, but our method provides the foundation for a turnkey solution in developing controllers for wearable robots,” said Shuzhen Luo, an assistant professor at Embry-Riddle Aeronautical University and first author on both studies, who previously worked as a postdoc in both Zhou's and Su's labs.
Typically, users must spend hours “training” their exoskeleton so that the technology knows how much force is needed and when to apply that force to help them walk, run, or climb stairs.
This new method incorporates both the exoskeleton controller and physical models of musculoskeletal dynamics, human-robot interaction, and muscle responses in a closed-loop simulation, allowing users to quickly put their exoskeleton to use, generate efficient, realistic data, and iteratively learn better control policies in simulation.
The unit is pre-programmed for immediate use, and the on-hardware controller can be updated if researchers make improvements through extensive simulations in the lab. Future prospects for the project include the development of individually customized controllers to assist with various activities of daily life.
“This research essentially makes science fiction a reality, enabling people to use less energy to perform a variety of tasks,” Su said.
For example, in tests with human subjects, researchers found that study participants expended 24.3% less metabolic energy when walking with a robotic exoskeleton compared to walking without an exoskeleton. Participants expended 13.1% less energy when running with the exoskeleton and 15.4% less energy when climbing stairs.
While the study focuses on researchers studying able-bodied individuals, the new method aims to help people with mobility impairments using assistive devices.
“Our framework may provide a generalizable and scalable strategy for the rapid development and widespread deployment of a range of assistive robots for both able-bodied and motor-impaired individuals,” Su says.
“We are in the early stages of testing the performance of this new method on robotic exoskeletons used by elderly people and people with neurological conditions such as cerebral palsy. We also hope to investigate how this method can be used to improve the performance of robotic prosthetic limbs.”
Funding: This research was supported by the National Science Foundation (grant numbers 1944655 and 2026622), the National Institute on Disability, Independent Living and Rehabilitation (grant number DRRP 90DPGE0019), the Switzer Research Fellowship Program of the Administration of Community Living, and the National Institutes of Health (grant number 1R01EB035404).
About this AI and Neurotechnology Research News
author: Derrick Raymond
sauce: New Jersey Institute of Technology
contact: Derrick Raymond – New Jersey Institute of Technology
image: This image is provided by New Jersey Institute of Technology
Original Research: The access is closed.
Xianlian Zhou et al. “Exoskeleton Assistance with Learning in Simulation, No Experiments Required” Nature
Abstract
Exoskeleton assistance without experimentation through simulation learning
Exoskeletons hold great potential for enhancing human motor performance. However, their development and widespread adoption is limited by the need for extensive human testing and handcrafted control laws. Here we present an experiment-free method for learning generic control policies in simulation.
Our simulation learning framework leverages dynamics-aware musculoskeletal and exoskeleton models and data-driven reinforcement learning to bridge the gap between simulation and reality without human experimentation.
The learned controller is then deployed into a custom hip exoskeleton to automatically generate assistance across a range of activities while reducing metabolic rates by 24.3%, 13.1%, and 15.4% for walking, running, and stair climbing, respectively.
Our framework may provide a generalizable and scalable strategy for the rapid development and widespread deployment of a range of assistive robots for both able-bodied and motor-impaired individuals.
