Actually, I've pondered something similar for a few years now.
Most suspensions react to the road. They read speed, and movement of the axles/suspension.
What you are describing, and what I've always wanted to build, would be a proactive suspension. While technically, it's still reactive, it would use cameras, radar, magic, whatever. the point is, it would react to road issues BEFORE the wheel actually struck them.
You could 'see' a pot hole and soften the suspension quickly.
you could 'see' a raised surface and soften it
you could see a speed bump, and absorb it.
My system design also calls for a learning system.
Just like how I assume the AWD tesla works... it senses what it's doing, what the result is, and learns from it.
You don't pre-program "hey this is a pothole" into the camera and say "soften the suspension"
nope. not one line of code.
What you do is program, ok, this is occurring (pot hole), try (something) it will (stiffen suspension). Then record the results. Then try something else next time.
record results.
when results move in a favorable direction, keep doing that thing, until they do not. Then use that value.
ta-da, self learning suspension.
I assume the AWD S will get more efficient over time, as it records the driving habits of 1,000s of cars. Driver input: 45% gas pedal, record speed, record acceleration, record energy usage per motor.
adjust energy levels per motor, slightly, over time, over every acceleration, forever adjusting and millisecond learning.
eventually, you'll produce the most efficient acceleration pattern.
All without knowing or caring what the actual efficiencies are. Because the car just learned the most efficient way of getting up to speed.
Of course that all goes to hell when you mash the go pedal
This allows accounting for subtle differences in all machinery... when you make 100,000 of something... some of them will be different... so why should they all have the same programming?
Observe, react, record, improve.
Lucky for you guys I'm more John Connor than SkyNet