This is one of those cool ideas that is much easier to grasp with some animated visuals. I haven't found one for this specific application, but adjacent work in using sound to move small objects is seen at https://www.youtube.com/watch?v=XpNbyfxxkWE , https://www.youtube.com/watch?v=3MXVSdXZzpc , and https://www.youtube.com/watch?v=tzWP-NL3Lck
Controlling the motion involves a lot of prediction and wave modeling, which should be a perfect application for high-performance computing.