Assuming actuators are expensive. The best approach must be to reduce the number of actuators to 1 main driver to cycle through columns plus 1 toggling actuator per row. (Further reduction into a complete XY-board will be too slow and complex). Move the memory to the mechanical domain.
That will reduce the number of moving parts to 1-2 per pixel, while still keeping the actuators complexity rather simple. The ballpoint pen isn't a bad idea, it just requires refinement. With some clever engineering i bet you could make that out of one single plastic part that can act both as a spring and a locking "click"-mechanism.
Similar to the Nist wheel driven display linked in the article [0] but invert it, such that the wheel rolls over a flat grid of miniaturized "ballpoint pens" to set them in the desired state. Rendering a full screen will take 1-2 seconds of rolling the wheel over all the columns. This should be an acceptable compromise.
That will reduce the number of moving parts to 1-2 per pixel, while still keeping the actuators complexity rather simple. The ballpoint pen isn't a bad idea, it just requires refinement. With some clever engineering i bet you could make that out of one single plastic part that can act both as a spring and a locking "click"-mechanism.
Similar to the Nist wheel driven display linked in the article [0] but invert it, such that the wheel rolls over a flat grid of miniaturized "ballpoint pens" to set them in the desired state. Rendering a full screen will take 1-2 seconds of rolling the wheel over all the columns. This should be an acceptable compromise.
[0] https://www.youtube.com/watch?v=sBBqdEnZPck