Myosuite simulates the interaction of human muscles, bones and joints. The findings could also flow into the Metaverse.
The AI platform includes matched, physiologically accurate musculoskeletal models, Meta researchers write in the company’s tech blog.
“The more intelligent an organism is, the more complex the motor behavior it can exhibit. So an important question to consider, then, is — what enables such complex decision-making and the motor control to execute those decisions?”, the researchers write.
MyoSuite is a set of musculoskeletal models and tasks that enable the application of ML to solve biomechanic control tasks. Such “control tasks” include complex finger movements: juggling a pen in the hand, turning a key, or manipulating two Baoding balls. The AI model is currently limited to the anatomy of human arms and hands.
Myosuite: foundation for advanced avatars?
The research aims to unravel the subtleties of sensorimotor control by simulating them as accurately as possible. According to Meta, the AI model developed works 4,000 times more efficiently than previous simulators, allowing far more muscles to be simulated simultaneously than before.
“We can train these models to perform complex movements like twirling a pen or rotating a key. This research could accelerate development of prosthetics, physical rehab, and surgery techniques,” said Meta CEO Mark Zuckerberg on the release of Myosuite. “It could also help us develop more realistic avatars for the metaverse.”
Meta has been researching so-called codec avatars for realistic immersive VR or AR conferences for many years. Zuckerberg sees simulated co-presence as the killer app for these future technologies.
Sensorimotor AI models could help to represent movements more realistically, regardless of whether they are detected by sensors or only vaguely inferred, as in inverse kinematics. This method estimates the spatial relationship of limbs based on limited data.
Metaverse research is human research
Findings at the intersection of neuroscience and motor control could also aid the development of neural interfaces. Meta is working on a wristband that intercepts muscle signals at the wrist and translates them into computer commands. In the future, such a device should make it possible to operate augmented reality headsets effortlessly and virtually invisibly.
In any case, Myosuite shows how far-reaching Meta’s research is. It ventures into medical and biological areas whose technical benefits only become apparent at second glance. At least if you assume they are designed for virtual reality and AR applications. Deceptively real avatars require precise knowledge of human anatomy. Corresponding basic research is therefore important.
More information about MyoSuite is available on the project’s official website. Benchmarks for AI researchers are available on Github. The project is the result of a collaboration between Meta AI and the University of Twente.