Fairly certain there is not enough information from one magnetic sensor to uniquely determine the position and orientation of the floating magnet. Once you actually measure the position and orientation of the levitated mass, its magnetic field, and the shapes of the supporting field. Then it is much easier. Do you think those rockets flipping quickly and then landing precisely upright were flown by seat of the pants algorithms? I expect it is “3D models with mass and precise locations”, a fair measure of “response functions”, “equations of motions – masses, velocities, accelerations, moments of inertia, etc” You only have 12 core observables to track – x y x position of the center of the magnet. And the cells of a 3×3 rotation matrix to bring it to the proper orientation.
Monitor those and simply use an objective function like the “total energy needed to rotate the disk and position it from its current position to the standard one. If the disk is not symmetric you can grab it with the field. So put “something” asymmetric on the disk to know up from down. east from west. etc. Any small magnets might work. You cannot easily grab a smooth magnet, unless you go to much higher sampling rates for observation and control.
I love your signals, but it is frustrating to see you just winging it. A little bit of data and statistics. Use cameras to monitor position (that is why you need some way to “see” which way it is pointing, and some way to grab it and exert force to rotate in 3D or lift or pull down.
Yes, you are close to the right field strengths and control signals and parts. But you have to measure, record and estimate the work needed (power times time) for each correction. You are flying the disk, not lifting it. Put yourself on the disk and it is easier to know what to do. Tilt five degrees down toward 23 degrees (clock wise from north)? You need a force to lift up at 23 degrees, and maybe one at 180+23 to lift up. The magnetic field works stronger over shorter distances, so lifting the nears part of the disk is easier than pulling the further one. When it gets level opposite sides are about equal. Get some coordinates and flight instructions. And keep track of energy., You have all to voltages, currents, charges, powers. You know or can measure the moments of inertia of the disk and anything attached.
Best wishes. You are really close. Me, I could not even see those little things you were soldering and can only see the mathematics and physics, sensors and data.
Put the disk on a thin panel in the right place. Then use the field to lift one side by exactly 3 mm, the put it down again. Repeat around the circumference and calibrate your currents. Get a separate 3 axis magnetic sensor and measure the fields. Strip down a 3D printer, set the fields at a given set of activation currents, and map the field so you can see it and know its values and fluctuations.