Every person moves in a different way. However, many exosuits - suits that support their user by reducing the load on the body when performing actions, among other things - work in one uniform way. Thus, the suits are not tailored to the specific way of movement of its wearer and therefore cannot optimally support their user. A new algorithm developed by the Wyss Institute for Biologically Inspired Engineering and the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) at US Harvard University makes it possible to tailor the operation of exosuits to their user and optimise the support provided by the suits.
Exosuits are used for a variety of purposes. For instance, there are examples of exosuits being used in industrial environments and warehouses to reduce the strain on workers while performing certain tasks. Project March is additionally working a on exosuit that enables patients with paraplegia to walk again. Harvard University's algorithm could be of interest for both applications.
Optimising control
"This new method is an effective and fast way to optimise the control of assistive wearable devices," explains Ye Ding, a postdoctoral fellow at SEAS and involved in the research. The algorithm Ding developed with other researchers is specifically designed to optimise the support exosuits provide their user at the hip. "Using this method, we have achieved a great improvement in the metabolic performance of wearers of a support device that forms an extension at the hip."
Incidentally, it has long been possible to tune the control of an exosuit to the specific movements of an individual user, although this does not happen with every exosuit. However, this is a manual and time-consuming process. "If you previously had three different users with assistive devices, you needed three different assistive strategies for this," said Myunghee Kim, postdoctoral fellow at SEAS and also involved in the research. "Finding the right control values for each wearer was a difficult and step-by-step process. This is not only because every person walks just differently, but also because the experiments required to find the right values are complicated and time-consuming."
Automatically finding the right settings
The new algorithm offers a solution and can significantly speed up this process. The algorithm helps find the right values in a short time to optimally support a user and minimise the energy consumed while walking. For this, the researchers use 'human-in-the-loop' optimisation, a method in which physiological signals from a wearer such as breathing are measured in real time. Based on these signals, the values used to control the exosuit are continuously adjusted automatically, with the algorithm telling the exosuit where and when to support the user. This process is carried out until the optimal settings are found.
By combining an exosuit with the algorithm, the researchers managed during experiments to reduce the metabolic cost - the amount of energy consumed to perform a given task - of walking by 17.4% compared to walking without an exosuit. This is a 60% improvement over the team's previous experiments without the algorithm.
Big impact on wearable robotic devices
"Optimisation and learning algorithms will have a major impact on future wearable robotic devices designed to support a wide range of behaviours," stated Scott Kuindersma, assistant professor of Engineering and Computer Science at SEAS. "These results show that optimising even simple controllers while walking can provide significant and individualised benefits to users. Extending these ideas to more expressive control strategies and people with both different needs and abilities is an exciting next step."
"With wearable robots like soft exosuits, it is critical to provide the right support at the right time so that they work synergistically with the wearer," says Conor Walsh, assistant professor of Engineering and Applied Science at the Wyss Institute. "With these online optimisation algorithms, systems can automatically learn how to do this in 20 minutes, maximising the benefit to the wearer."
In the coming period, the researchers will focus on applying the algorithm to more complex devices that support multiple joints simultaneously, such as the hip and ankles.
Author: Wouter Hoeffnagel
Source: Wyss Institute for Biologically Inspired Engineering at Harvard University
Source photo: Pixabay / www_slon_pics
