I envision using cam levered suction cups to hold on rotary and linear sliders that had touch screen sensitive rubber tips. One could go as far to 3d print arbitrary interaction devices that could get attached to the face of the touch screen. You can use the multitouch sensor w/o the screen but still be able to configure arbitrary devices to go on the front.
I even had a design for joystick. Lots of analog opportunities when you have something like a back illuminated camera or a touch controller that can sense areas. You could also serially transfer data from the device to the touch screen, either using physical touches or electrically simulated touches.
I'd be surprised if the concept weren't already patented, though, just because the idea of a generalized capacitive control surface seems fairly obvious, and the patent office doesn't really apply an "obviousness" test. What definitely surprises me is that, patented or not, I can't just go out and buy these sorts of controls.
We would then be able to receive that pattern and understand how the knob is rotated.
What’s the resolution of todays touchscreen?
But for simple things, if possible without affecting the other parts of the screen, it'd be amazing to have a broadly supported, low bandwidth standard.
This might be as simple as "Here's my encoded position * frequency of sampling", but for a general interface you'd want something adaptable.
What if there were two dials on the control? Two dials and three buttons? One dial, four buttons, and a joystick?