Researchers at the University of California San Diego have developed BIGE, a generative AI model aimed at preventing injuries in athletes. By integrating biomechanical constraints, BIGE provides realistic motion patterns for training and rehabilitation.
Researchers at the University of California San Diego have developed a groundbreaking generative AI model named BIGE (Biomechanics-informed GenAI for Exercise Science) aimed at preventing injuries among athletes and aiding in their rehabilitation.
Utilizing athlete movements and biomechanical constraints — such as the muscle force limits — BIGE generates videos of realistic and safe exercise motions that athletes can follow to minimize the risk of injury.
The model can also suggest movements allowing injured athletes to continue training without exacerbating their injuries, potentially revolutionizing both sports training and rehabilitation practices.
“This approach is going to be the future,” co-senior author Andrew McCulloch, a distinguished professor in the Shu Chien-Gene Lay Department of Bioengineering at UC San Diego, said in a news release.
BIGE stands out as the only model combining generative AI with realistic biomechanics, contrasting with many existing AI models that frequently produce anatomically inconsistent results. Non-AI methodologies, on the other hand, often require an unfeasible amount of computational power.
The research team trained BIGE using motion-capture videos of individuals performing squats, translating these movements onto 3D skeletal models to compute physical forces. This enables the AI to generate more lifelike exercise motions, offering athletes safer training protocols and improved performance outcomes.
Accepting that the technology is in its early stages, the team sees significant potential for expansion beyond squats. Next steps include adapting the model for various exercises and personalizing it for individual users.
“This methodology could be used by anyone,” added co-senior author Rose Yu, a professor in the UC San Diego Department of Computer Science and Engineering.
The research team also envisions applications beyond the sports world. For instance, BIGE could assess fall risks in the elderly, potentially benefiting a broader demographic and marking a significant stride in public health and safety.
The team recently presented their research at the Learning for Dynamics & Control Conference at the University of Michigan in Ann Arbor.