Adam Kiefer is working on two projects that use data science to help enhance athlete performance, prevent injuries, and assist with injury recovery through AI and behavioral dynamics that analyze, measure, and model complex human health and performance. In the first project, the analysis includes using a digital twin — a digital model of a person connected to the individual via data and artificial intelligence (often in real-time) — for simulations and training. These digital twins allow Keifer’s team to model physical systems and see how they change, adapt, evolve, and even potentially break down under different circumstances. Digitally modeling a human being is a complex process. It involves ingesting historical, pre-recorded data, using existing behavioral models to inform a fuzzy inference machine learning model, and pulling data from portable digital eye tracking from the athlete (to gain insight into an athlete’s visual tracking and thoughts while in play) and from the sidelines (to inform pose estimation and action recognition). Based on the model trajectory, hyper-parameterization, and raw data, the system selects a set of data samples from their libraries and plays out the data as a simulated behavior. As the data has time-dependent structures, digital twin scenarios can be played out over time, allowing researchers to inform coaches and athletes about how various interventions may impact performance. Kiefer’s second project, funded by the NIH, provides smart, dynamic, immersive virtual reality assessment to athletes returning to their sport after injury. In this project, Keifer’s team has designed a virtual reality system that adapts to the athlete’s capability through graded tactical challenges as the athlete trains and improves. It includes using peripheral data, such as biomechanics, neurophysiology, eye tracking, and pose estimations and delivers trial-by-trial difficulty, based on the athlete’s performance. To improve its use in training, the team incorporated a second-order differential equation behavioral model and genetic fuzzy logic inference, where they can predict athlete decisions up to one second in advance, so AI defenders can be informed to force the athlete into more adaptive behaviors. The goal is to help athletes train and return to peak performance more rapidly after injuries.
Adam Kiefer, Assistant Professor, Department of Exercise and Sport Science; Co-Director, STAR Heel Performance Laboratory
Department: Exercise and Sport Science | Faculty Profile
Featured on: January 26, 2023 (Event Page)
Session Title: Advancing Education, Training, and Care (Event Recap)
Tools, Information, and Resources:
- Amazon Web Services (AWS): Amazon Web Services offers reliable, scalable, and inexpensive cloud computing services.
- Tobii: Tobii is on a mission to improve the world with our eye tracking and attention computing technology that understands human attention and intent.
- Biomechanics in Motion | Virtual Reality: In this video, researchers from the UNC Motion Science Institute will discuss the use of virtual reality to assess on-field movement strategies.
- Cloud-enabled Computer Vision Platform for Automated Assessment of Basketball Shooting Performance<: Basketball athletes and coaches know the importance of vision and technique for shooting success, but they have been unable to quickly and objectively assess these behaviors to inform actionable training plans. Researchers in the STAR Heel Performance Laboratory at the University of North Carolina at Chapel Hill worked with Amazon Web Services (AWS) and an eye tracking hardware partner, Tobii Pro, to develop and deploy an end-to-end automated eye and movement tracking basketball shooting assessment system to test, analyze, and profile athletes for individualized training to optimize vision and improve performance.