The gym is crowded with individuals performing squats incorrectly. Knees collapsing inward, backs curving, hips misaligning. Some manage to evade issues. Others find themselves sidelined for an extended period, questioning where they went astray. Now, researchers at UC San Diego have developed an AI system that may help prevent athletes from visiting the trainer by creating customized videos that demonstrate precisely how to execute movements.
This model, referred to as BIGE (Biomechanics-informed GenAI for Exercise Science), does not simply produce standard exercise footage. It integrates generative AI with genuine biomechanical limitations, such as the force a muscle can exert or the safe angles for joints. When provided with motion-capture data of an individual squatting, it generates videos of movements designed to minimize injury risk or facilitate recovery post-injury.
How This Stands Apart From Fitness Applications
Most AI frameworks aimed at simulating human motions face a challenge. They may make it appear accurate, but the underlying physics could be completely erroneous. A person might seem to execute a perfect squat, yet the forces on their knee joints could be incorrect. To the researchers’ knowledge, BIGE is the sole model that effectively merges generative AI with realistic biomechanics. The alternative methods, which factor in physics without employing generative AI, demand so much computational power that they are essentially impractical outside of research facilities.
In training the system, the team utilized motion-capture recordings of individuals performing squats, subsequently mapping those actions onto three-dimensional skeletal representations. By calculating the involved forces, they generated motions that are not only visually convincing but also physically feasible. The yellow trajectories in their demonstration videos illustrate hip joint movement throughout the entire squat cycle, with BIGE’s output appearing smoother and more natural compared to existing foundational models.
Andrew McCulloch, distinguished professor in the Shu Chien-Gene Lay Department of Bioengineering at UC San Diego and lead author of the work, forecasts: “This approach is going to be the future.”
Expanding Beyond Squats and Into Rehabilitation
At present, BIGE specializes in squats. Upcoming goals include broadening its scope to encompass additional movements and customizing the models for individual users. The potential uses extend far beyond athletes. Rose Yu, a professor in UC San Diego’s Department of Computer Science and Engineering and another lead author, highlights that the methodology has extensive applicability. For instance, it could assist in evaluating fall risk among elderly groups.
Yu adds: “This methodology could be employed by anyone.”
The team showcased their findings at the Learning for Dynamics & Control Conference at the University of Michigan in Ann Arbor. For athletes recovering from injuries, the system could create movements that enable them to continue training while safeguarding injured tissue. For those aiming to prevent injuries initially, it provides a means to visualize what proper form specifically entails for their unique bodies.
It’s one thing to be informed that your squat technique requires improvement. It’s another to view a video crafted exclusively for you, demonstrating the precise adjustments your joints and muscles need to implement. Whether this will ultimately keep individuals free from injuries remains to be seen, but at least the physics align.
Journal: Proceedings of Machine Learning Research, Vol. 283, pp. 1243-1256
Conference: 7th Annual Learning for Dynamics & Control Conference
Full paper: https://proceedings.mlr.press/v283/maheshwari25a.html
If our reporting has informed or inspired you, please consider making a donation. Every contribution, no matter the size, empowers us to continue delivering accurate, engaging, and trustworthy science and medical news. Independent journalism requires time, effort, and resources—your support ensures we can keep uncovering the stories that matter most to you.
Join us in making knowledge accessible and impactful. Thank you for standing with us!