In sports such as gymnastics, the smallest body movements can have a gargantuan effect on performance. Even the slightest twitch can make the difference between a top finish and a devastating loss. In the perpetual quest to push toward new physical limits, world-class athletes are turning to advanced technologies that track fluid dynamics and body motion. These technologies provide data that informs small physical adjustments that the athlete can make in training and even during competition. To adequately capture a realistic simulation, however, requires advanced technology — and a significant amount of processing power. A skilled engineer can generate a complex 3D model of an athlete using CAD software, for example, but it might take days to create such a model. Coaches and athletes often need feedback in just minutes. This time delay has led to increasing use of 3D scanning technologies, which can produce an accurate point cloud of millions of data points in minutes. A streamlined simulation A solution that combines 3D simulation software and artificial intelligence (AI) to efficiently capture data and produce full-body, 3D avatars is helping world-class athletes to push the bounds of human performance. Developed by Falcon Pursuit, the Athlete-Avatar system reduces barriers to adoption and the time required for the system to perform computational fluid dynamics — while maintaining accuracy. Using data provided by the system, engineers can quickly generate a highly accurate digital model for use in fluid dynamics software. Falcon Pursuit’s 3D simulation software interface. After completion of a 3D scan, the system operator selects a fully rigged base avatar, along with a full structure of muscles and bones from the software library. The Athlete-Avatar platform combines 3D simulation software and AI to produce full-body, 3D avatars that deliver feedback to inform users of dynamic motion. Courtesy of Falcon Pursuit. The system generates a model in the form of an avatar of an athlete’s body and body movements. The system provides coaches and athletes information that they can access in just a matter of minutes. The comprehensive solution incorporates 3D imaging, point-cloud data, and powerful compute to deliver feedback that ultimately gives athletes a chance to earn a better shot at a first-place finish. The Athlete-Avatar system uses point-cloud data and a library of high-fidelity avatars to analyze how athletes move in 3D space, providing accurate computational monitoring of air and fluid motion. The process begins by making a 3D scan of an athlete, which takes place inside the company’s AVA kiosk and completes in just minutes. The kiosk comprises seven Structure Core 3D sensors from Occipital that each feature stereo infrared cameras, an RGB camera with 1280 × 960 resolution and 160° FOV, a NU3000 multicore processor from Inuitive, and 6-axis inertial measurement units. When the 3D scan is complete, the system operator then selects a fully rigged base avatar from the software’s library that has a full structure of muscles and bones. The point cloud from the scan wraps around the selected avatar. The result is a creation that resembles a Silver Surfer-like model of the athlete. After the operator saves the avatar, the software removes the point cloud because tracking each individual piece of data is no longer required. This step greatly reduces processing requirements, and as a result it also reduces processing time. Meanwhile, the software keeps the complex shape and accurate movements of the athlete. By providing more flexibility, as well as information on drag, this approach ensures that the avatar moves more elegantly than avatars created by scans or STL (stereo- lithography) files. The approach also reserves time for processing real-time feedback about performance. Creating avatars with 3D cameras During the scanning process, sensors and cameras mounted on robot arms move around the athlete’s body. The system produces digital renderings that offer precise modeling of the athlete’s form, figure, and 3D motion. The system uses four UI-5581LE board-level cameras from IDS Imaging Development Systems. The cameras feature the 5-MP MT9P006STC color CMOS image sensor from onsemi that delivers a resolution of 2560 × 1920, and they use a proprietary lens that minimizes the “flat view” effect that occurs when working with extremely close subjects. The Athlete-Avatar system uses point-cloud data and a library of high-fidelity avatars to analyze how athletes move in 3D space, providing accurate computational monitoring of air and fluid motion. Inside the AVA kiosk, a 3D scan is performed in minutes. The kiosk comprises 3D sensors that use stereo infrared cameras, as well as an RGB camera, a multicore processor, and 6-axis inertial measurement units. Courtesy of Falcon Pursuit. Users of the system can configure the avatars into hypothetical stances to test the athlete’s best performance methods, or users can re-create the athlete’s performance after it has been captured on conventional 2D video. The use of an avatar delivers a high level of accuracy and flexibility to analyzing drag and other key indicators of performance in 3D space. Falcon Pursuit’s avatar-rendering engine is powered by a Nuvo-7000 PC from Neousys Technology that was customized by CoastIPC. This fanless industrial PC features an i7-8700 TE Intel Core processor, 16 GB of memory, and a GT 730 GPU from NVIDIA. It provides the necessary power and cooling requirements for the system to produce high-fidelity 3D digital scans of an athlete in the short amount of time required by athletes and coaches. Before winning a gold medal in 2021, an Olympic athlete used Falcon Pursuit’s Athlete-Avatar tool in her training. The technology enabled her and her coach to extract information about her body rotations and skeletal positions, and it generated a visual display within minutes of a completed training session. The system performed AI analytics of body motion repetitions in multiple exercises, including balance beam. Courtesy of Falcon Pursuit. Combined with machine vision and AI processing, the software can match 3D movements of the avatar to any 2D video of the athlete for analysis. The software also allows users to manipulate the avatar to analyze how changes in movement could affect performance. Going for gold The Athlete-Avatar technology is deployed by coaches and athletes across a range of different sports, in addition to gymnastics. Swimmers, cyclists, and surfers are also benefitting from using the technology. Prior to winning an Olympic gold medal in gymnastics in 2021, one female athlete worked with coach Jess Graba and Falcon Pursuit during training. Throughout a day, she was scanned in the AVA kiosk so that the system could perform advanced AI body motion analytics during beam, floor exercise, uneven bar, and vault training. “Using the system, we applied AI analytics to training or broadcast video [of the athlete] to extract full body rotations and skeletal positions to graphically display for the athlete to review minutes after completing any routine,” said Falcon Pursuit CTO Jay White. The athlete continues to use the data she received from the technology. In partnership with Red Bull, Falcon Pursuit used its technology to also generate AI analytics for Olympic surfers Caroline Marks and Kolohe Andino, as well as in cooperation with Dalilah Muhammad, an Olympic track and field athlete who specializes in hurdles. According to Falcon Pursuit, the system could also be used for mechanical, thermodynamic, and hydrodynamic simulation, allowing the potential expansion of the technology into new and exciting applications. Note: The methods described in this article are patented under U.S. patent numbers 9,797,802 and 10,648,883. News editor Jake Saltzman contributed to this piece.