[spaceandshit]

Convince regulators that advanced control techniques can be used on flight software (model-based predictive control, optimal control, etc.)

We know how to do these, maybe not as safe as we can with simpler methods today, but convincing the public and regulators will always be the challenge in aviation/aerospace.

Probably the only realistic (software related) method in today's world without a materials/battery improvement, but we won't get there for a while.

[wenc]

So I don't know the first thing about aircraft control, but I'm a control engineer by training.

What control scheme do aircraft use today? I always thought it was a form of MPC (well GPC = generalized predictive control) with a state estimator like a linear Kalman filter.

[spaceandshit]

Are you familiar with DO-178? Basically outputs have to be deterministic.. so you can't really use advanced control techniques.

I haven't worked on flight control (but control on specific aircraft systems, which are at the same level), and they're all basically PID control thrown together.

[wenc]

Ah no, aircraft control is outside of my province, so this is new knowledge.

Advanced control can be deterministic -- through parameterization. Explicit MPC [1] does this. It pre-computes the quadratic optimization problem offline, so all the solutions are enumerated and stored, and are verifiable. However, as far as I know no one has implemented this commercially.

Interesting that it's all PID. Not surprised though because PID is tried-and-true (but even so, has a lot of tiny nuances which takes years to understand).

[1] https://www.mathworks.com/help/mpc/explicit-mpc-design.html

[spaceandshit]

Thanks for the link, I was not familiar. I'm not sure if any current commercial systems use these methods. I know the defense industry uses a lot of advanced techniques because they're their own regulators.