This is fun experiment done by graduate students of Urban Champaign.
They point out a 1989 paper [E. H. Brandt, “Levitation in physics,” Science, vol. 243, no. 4889, pp. 349–355] to illustrate how such system was long known to scientific community. However, getting ball into complex trajectory is a tricky business. I agree. The control strategy being implemented is a feedback mechanism with implementation of extended Kalman filter for state estimation. They track the object with stereo vision camera and use classical fluid dynamics model for the air jet being mounted with two degrees of freedom.
Whenever possible, have a look at their [paper].