My controls education…

During my time in Masters school I was given lots of simple models for which I was asked to design controllers. As an undergrad I was given several examples of systems to develop simple rigid body models for and then design a controller. All in all, when I graduated I found my education with regard to system modeling to be lacking.
My undergrad is in Mechanical Engineering and we did plenty of rigid body models of multiple degree of freedom systems. My Dynamics class was probably my best with regard to system modeling. However, this completely ignores the modeling of complex sensors and actuators.

An education in modeling

Modeling is largely a matter of judgement. Judgement based on experience but also based on performance. In LaserCom applications we are attempting to control line of sight (LOS) pointing to less than 10 micro-radian level with milli-radian (or larger) level disturbances at high frequencies. As a result we have to model not only 1st order effects but 2nd and 3rd order effects. If we were attempting to control the LOS to a milli-radian level or hundreds of micro-radians then 1st order effects would be adequate.

My controls education recommendation

My recommendation would be to scrap the typical undergrad controls course and turn it into a semester long project. The projects would start with an assessment of desired system performance. That performance becomes a way to derive requirements. The requirements become a means of determining whether the model needs to include 1st, 2nd or 3rd order effects. Actuators and sensors are modeled in successive stages of complexity. After the system, sensor, and actuator models are created then a full plant model can be formed – it will probably be at least 6th order.
The full plant model can then be approximated in order to design a controller. Then that controller can be tested against the approximate plant and the full plant. Root locus, pole placement, and PID controllers can be explored during this design phase.
Obviously the design of the controller will benefit from open loop, closed loop, and disturbance rejection bode plots. Stability can be assessed using the Routh stability criteria, Nyquist, Nichols, etc.

Once the controller performs well on the full plant quantization, saturation, etc can be explored.

At the end of the semester the students have the tools and they’ve used the tools.