deatbeat controller

Techniques that require pole-zero cancellation are generally not robust.

fred

Unless you can eyeball the system and know what pole locations are safe

-- and that doesn't really count.

Because it generally requires absurdly high gains, and loop closing frequencies well above the points where the plant model is no longer reliable.

You may be able to make a very high-order deadbeat controller that's robust, by padding it's response enough to bring the loop closure frequencies down -- this would be an interesting research topic. Keeping it deadbeat in real life would undoubtedly require an adaptive controller.