I graduated from Delft in 1985 as an aeronautical control engineer on two separate theses, because of a tiff between me and the distinguished professor Gerlach, who was a nasty piece of work. And maybe I was just a little bit cocky.
My first thesis was on an asymmetrical autopilot ( i.e. for left / right turns ), for the DHC-3 "Beaver" laboratory aircraft which had recently been fully identified in flight. The thesis was in Dutch, which was the standard at the time.
I used an interesting feedforward take on turn coordination when rolling into a turn, and a charming cascaded setup for the localiser approach, starting from an innermost loop for the roll angle, around that a loop for the course angle, and aroudn that again a loop over the offset from the runway centerline. This gave good insight on the gains and time constants for each loop, and equally importatn, it allowed natural settable limiters on every level in the nested loops : first a limit on roll rate, then a limit on the roll angle ( set to 30° ), and finally a limit on the intercept course towards the runway centerline, to prevent circling.
There was no love lost between me and my professor, and he forced me to do a second thesis, this time in English, on the symmetrical ( up / down ) control.
I used a cascaded controller again, starting from an innermost loop for the angle of attack α, right up to altitude hold and ILS approach capture. In the system equations, I used the flight path angle γ instead of the pitch attitude θ as a state element ( and as a step in the nested, cascaded controller ), because this gave a very elegant inner loop for the ILS glide path and altitude hold intercept.
Like in the asymmetric controller, every step in the cascade was a setpoint for an inner loop, which makes it a natural place for a limiter, starting with a straightforward limit on the commanded angle of attack which automatically prevented ever commanding a stall.
In the symmetric system state I replaced the height h by the energy difference ( u + g / V . h ). Unlike h proper, this energy state is completely decoupled from all other modes, including the phugoid. It is directly coupled as a first order system to the throttle, which is now also decoupled from the phugoid. This is why I termed the fifth mode the "energy mode".
It is a known fact that the throttle is an energy control, not an airspeed or climb angle control. On approach, the energy management and in particular the trading between airspeed excess and height relative to the glide slope becomes fully automatic if feedback of altitude and airspeed is replaced by feedback of just the energy excess. When coming in slow and high or fast and low, there is no unnecessary throttle activity at all. In the root locus, feedback of the energy on the throttle moves only the single pole of the energy mode.
The professor told me: "that's not how we do it". Then he offered me a job at the NLR ( Dutch National Aerospace Laboratory ), where he was chairman of the board, in addition to his professorship in Delft. I politely declined.
PDF ' s :
Beaver autopilot symmetrical, 1985
Beaver autopilot asymmetrical, 1984