During my study, I discovered that PID (Proportional Integral Derivative) is also commonly used for SISO and linear models. As a result, I'd like to understand about the limitations of alternative controllers that necessitate model predictive control. I'd appreciate it if someone could recommend some papers to help me learn this.
A concise yet very excellent intro to MPC:
https://web.stanford.edu/class/archive/ee/ee392m/ee392m.1056/Lecture14_MPC.pdf
Notice if you do care about system constraints for your control application, and your model (to be controlled) is not highly nonlinear, then MPC is recommended over PID.
Hello Amrit,
If you intend to use Model Predictive Control, you should be willing to dive into control theory a little. Unlike PID control (for linear SISO systems, at least), applying MPC to your problem requires a fair bit of understanding of the underlying principles. The following publication provides a fine introduction to (nonlinear) Model Predictive Control: https://www.researchgate.net/publication/215530586_An_Introduction_to_Nonlinear_Model_Predictive_Control
Metaphorically speaking: Applying MPC is like driving by looking through the front window, whereas classical feedback control only reacts to deviations at the current time instance.
Advantages of (N)MPC in a nutshell:
- "foresightful" action
- ability to take state and control constraints into account
- potentially applicable to highly nonlinear system dynamics
Disadvantages:
- sound prediction model required
- high computational effort
- no guaranteed runtimes (in general)
- no guaranteed closed-loop stability (in general)
Best regards
Stefan
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning