On August 4 we took Squat to SquashBusters in Boston — a nonprofit that mentors underprivileged youth in both squash and academics. Ishaan Mehra ’27, Daniel Peregudov ’27, Milo DuBois ’28, Kanishk Venna ’29, and Ariyana Mehra ’30 made the trip.
We walked their players through how the robot actually works. Two high-speed cameras record at over 100fps, and a neural network finds the ball in every frame. We triangulate its position into real 3D court coordinates, then run that through a physics model tuned for squash — drag on the ball, and how it comes off the wall and the floor. What comes out is a prediction of where the ball is headed before it gets there. Mecanum wheels let Squat drive in any direction without turning, so it can slide into position and play the return. All of it runs onboard.
Afterward we set the robot up courtside and recorded their session, capturing rallies from Squat’s own point of view. That footage becomes training data for our ball-tracking models: real players, real pace, real lighting — exactly what the system needs to learn from. Every rally sharpens the next version.
Thanks to SquashBusters for having us. We’ll be back as Squat keeps developing.