Researchers from EPFL and UNIGE have created an advanced AI model that breaks new ground in understanding the intricate movements of the human hand. This innovation, which won the MyoChallenge at NeurIPS 2022, holds promise for revolutionizing neuroprosthetics.
In a significant leap for neuroscience and biomedical engineering, an AI model developed by researchers from École polytechnique fédérale de Lausanne (EPFL) and the University of Geneva (UNIGE) in Geneva has achieved a remarkable breakthrough in modeling the complex movements of the human hand. This novel approach not only triumphed in the prestigious MyoChallenge at the NeurIPS conference in 2022 but also sets the stage for revolutionary advancements in neuroprosthetics and our comprehension of motor functions.
Led by Alexander Mathis, a tenure track assistant professor at EPFL, the research team utilized an innovative machine learning strategy that merged curriculum-based reinforcement learning with highly detailed biomechanical simulations.
“What excites me most about this research is that we’re diving deep into the core principles of human motor control — something that’s been a mystery for so long. We’re not just building models; we’re uncovering the fundamental mechanics of how the brain and muscles work together,” Mathis said in a news release.
Tackling the Baoding Balls Challenge
The challenge organized by Meta required participants to create an AI capable of manipulating two Baoding balls using 39 muscles in a coordinated manner.
Three graduate students — Alberto Chiappa from Mathis’ group and Pablo Tano and Nisheet Patel from Alexandre Pouget’s group at UNIGE—excelled in this daunting task. Their model attained a 100% success rate in the initial phase of the competition and continued to outperform competitors in increasingly difficult conditions.
“To overcome the limitations of current machine learning models, we applied a method called curriculum learning. After 32 stages and nearly 400 hours of training, we successfully trained a neural network to accurately control a realistic model of the human hand,” added Chiappa.
Implications for Neuroprosthetics
A remarkable aspect of this AI model is its ability to identify and replicate basic, repeatable movement patterns or motor primitives. These findings could provide critical insights into how the brain learns and executes complex tasks, influencing the development of advanced neuroprosthetics.
Silvestro Micera, a leading researcher in neuroprosthetics at EPFL’s Neuro X Institute and collaborator with Mathis, underscores the significance of this research.
“What we really miss right now is a deeper understanding of how finger movement and grasping motor control are achieved. This work goes exactly in this very important direction,” added Micera. “We know how important it is to connect the prosthesis to the nervous system, and this research gives us a solid scientific foundation that reinforces our strategy.”
A Path Forward
The implications of this research stretch far beyond the bounds of the laboratory.
By providing a dynamic, anatomically accurate model of hand movement inspired by human motor learning, this AI-driven approach may soon enable the creation of prosthetic devices that offer seamless, natural control for users with limb loss or paralysis.
Additionally, it paves the way for further exploration into the brain’s intricate mechanisms that govern movement, promising broader applications in both healthcare and robotics.
The results of this innovative research have been published in the journal Neuron, offering a glimpse into a future where technology and biology interact in ever-closer harmony to solve some of the most pressing challenges in human health and functioning.