A Look Underneath the Hood: How Driveline Sport Science Collects, Processes, and Analyzes 1000’s of Athletes’ Biomechanics Information
Driveline Baseball has lengthy been baseball’s utilized biomechanics chief. Since 2008, when founder Kyle Boddy began constructing his first movement seize lab within the aisles of Residence Depot, Driveline has leveraged the worth of objectively quantifying human motion.
The lab has improved by leaps and bounds since then. Driveline R&D has iteratively refined the expertise and underlying techniques in our movement seize ecosystem. Our present setup, The Launchpad, options simultaneous markered and markerless movement seize information in addition to a wealth of further monitoring applied sciences like Rapsodo, Trackman, HitTrax, and Pulse. Gone are the times of utilizing hacked Wiimotes to measure biomechanics (true story).
And that lab has enabled us to be taught numerous issues. We’ve studied the consequences of weighted baseballs on pitching mechanics, validated biomechanics applied sciences, and leveraged biomechanics to drive our coaching protocols for each hitters and pitchers.
Our Biomechanics Group has written and talked extensively on each the Driveline Weblog and Driveline R&D podcast on what we do with biomechanics and accent information as soon as we get it. Some wonderful examples embrace our ahead dynamics and computed muscle management simulations. Much less generally talked about is how we get to the purpose the place we’re prepared to research our information. How will we go from uncooked 3D coordinates to kinematic and kinetic information that analysts, coaches, and different R&D workers can use? Moreover, how will we mix information from so many sources to reinforce our insights with out overtaxing workers??
Within the lab, we’re at all times operating a handful of various applied sciences. So how will we handle all of it? As anybody who’s interfaced with a number of baseball applied sciences earlier than is aware of all too nicely, every further device is one other job for the coaches or participant growth workers. Because the tech stacks up, the workload shortly turns into overwhelming.
Until…
Every further toy is one other job except you construct the underlying techniques to cut back the quantity of human oversight wanted. How will we effectively course of biomechanics information and stream it into our broader participant growth ecosystem?
That’s what we’ll speak about right here.
Much less Time Amassing; Extra Time Creating
Despite the fact that a number of applied sciences could also be operating for any given particular person movement seize evaluation, customized software program options enable the Launchpad to be run by one individual – although we nonetheless want one other set of fingers to feed the pitching machine for hitting assessments.
Proprietary Driveline software program syncs our Edgertronic cameras to any Stalker radar gun enabling automated triggering when a pitch is thrown. This eliminates the necessity for handbook Edgertronic operation. Edgertronic clips are then synched to athletes’ TRAQ profiles and fed into our inside markerless movement seize pipeline.
Rapsodo and Trackman ball monitoring information are mechanically uploaded to our inside databases. Visual3D and Python processing pipelines add each pitch/swing (markered or markerless) to the identical information warehouse. Joint kinetics, kinematics, energetics, and power plate information are matched on a per-pitch degree whereas session and player-level tables present the required metadata for customized information choice and filtering as wanted. Customized information constancy checks filter out any troublesome inconsistencies that bumped into points alongside the way in which.
Ball monitoring and full sign biomechanics information get paired on a per-pitch degree. This whole course of, from inserting markers to producing customized reporting that populates in an athlete’s Traq profile, happens inside 24 hours of preliminary information assortment. If now we have a excessive precedence evaluation that requires an expedited turnaround time, we will manually run our processing pipelines to generate the required deliverables forward of the next-day turnaround.
All of those automations unlock R&D workers to do what they do greatest: help coaches and gamers, drive additional insights, and regularly enhance the underlying techniques.
Markerless and Markered Movement Seize Processing
Since Driveline’s first movement seize processing pipeline over a decade in the past, now we have been engaged on making biomechanics information processing as clean as doable. Immediately, we’re on the sixth iteration of our processing pipeline (v6) and the quantity of human oversight wanted to take information from evaluation to relevant perception has by no means been decrease.
We began making an attempt to maintain monitor of what number of labor hours go right into a single mocap session again in 2019. Again then, we had been turning round reviews in ~7 days and it was taking 5+ mixed labor hours per movement seize session. Immediately, we do extra with that biomechanics information than we ever have and through peak volumes it takes us lower than 90 minutes of labor per session from finish to finish. This summer season, there have been over 500 movement seize classes within the Launchpad and it solely value us ~600 labor hours.
However what will get us enthusiastic about V6?
with out making a gift of the key sauce…
Up to date Pose Estimation & Skeleton Mannequin
Visual3D has been the main research-grade biomechanics processing platform for a number of a long time. We really feel no different platform affords comparable choices in relation to pose estimation and linked-segment modeling. It’s for these causes that we funnel all of our markered and markerless movement seize information by means of Visual3D for foundational pose estimation, inverse kinematics, and inverse dynamics calculations earlier than transferring to Python and R for extra intensive and application-specific sign processing.
With V6, now we have used Visual3D to additional advantageous tune our linked-segment mannequin in order that it extra carefully captures how our athletes transfer. Our up to date mannequin is much less constrained, permitting segments on the distal finish of enormous vary of movement joints, such because the shoulders and hips, to maneuver freely relative to their neighboring phase. Beforehand, constraining segments on opposing sides of those joints generally triggered pose estimation issues and fewer correct kinematic and kinetic profiles.
Vitality Circulate + Power Plates
V6 homes two new information courses along with the standard kinematic and kinetic information: energetics and power plates. Energy flow analyses enable us to see when, the place, and the way a lot kinetic power is transferring between physique segments. Power plates inform us how our athletes are interfacing with the bottom. Each of those information sources assist present a extra full image of how athletes transfer throughout every pitch/swing.
Uncooked & Processed Information
V6 homes all our unfiltered, unprocessed movement seize information separate from the processed information. Any worker or shopper can entry it and carry out customized analyses for their very own initiatives. If we wish to check totally different filtering settings? Cool. If we wish to check out a unique set of mannequin assumptions? No downside. If we wish to discover a brand new metric derived from international kinematic/kinetic information as a substitute of counting on typical coordinate system and rotation sequence conventions? We are able to.
The place We’re Going
Much less invasive information assortment means bigger samples. With these bigger samples, we will begin to determine, on a person and group degree, what actually issues. Paired markerless information with live-AB outcomes drill proper right down to what get’s hitters out. Markerless information paired with ball monitoring and per-play outcomes permits us to research baseball at a degree that nearly didn’t appear doable only a few years in the past. Need to quantify how a pitcher’s deception influences an anticipated run worth of a pitch? We are able to. Need to measure how a batter adjusts to an off-speed pitch thrown in a hitter’s rely? That’s doable, too.
Much less labor intensive information assortment additionally means extra consideration to different applied sciences to proceed constructing our paired databases. Probably the most highly effective insights sometimes come on the intersection of sport science sub-disciplines. Much less labor intensive permits us to stay there, focusing extra on the interaction of method and efficiency.
Higher techniques automation and group tightens the suggestions loop between movement seize and actionable perception. No extra dedicating apply time to getting markered up and transported to some separate biomechanics lab to get a baseline or a retest mocap performed. Now we will do it with minimal disruption to the apply plan, and with out killing R&D workers with labor-heavy information cleansing or processing. All so we will hold digging into enhance on-field efficiency.
Written by Dr. Kyle Wasserberger – Sports activities Scientist
We may help you along with your biomechanics information. Be taught extra about our Biomechanics Providers right here
