Frequency Domain Tuning Method for Unconstrained Linear Output Feedback Model Predictive Control

Tuning Guidelines for Model-Predictive Control

“…However, the worst controlled variable responses were obtained with the guidelines suggested in ref which has the lowest controlled-variable weight to input rate of change weight ratios of 87/49,000 and 58/69,000. As expected from the controlled variable responses, the controller actions were initially most aggressive when the tuning guidelines proposed in refs and were used and were most sluggish when the guidelines in ref were employed.…”

“…Kokate et al 36 suggested setting ρ to a small value such as 0.1 in DMC, which outperformed a well-tuned proportional−integral−derivative (PID) controller in terms of overshoot, response smoothness, and robustness. For a state-space implementation of MPC, Burgos et al 62 suggested setting Q = C T C and R = ρI, and choosing a very small value for ρ. This approach was applied to a quadrotor reference tracking problem and yielded satisfactory set point tracking.…”

“…Input and output responses of the distillation column under MPC tuned with the four tuning approaches (Shridhar and Cooper, 18 Olesen et al, 46 Burgos et al, 62 and Waschl et al 73 ) when a step change was made in the x D set point. 9 and in the presence of the process-model mismatch.…”

“…Figures and compare the distillation column input and output closed-loop responses obtained under MPC with the four sets of tunable parameter values (Table ). In the case of a step change (from 70 to 90 at t = 0) in the x D set point (Figure ), the best responses of both controlled variables were achieved using the tuning guidelines proposed in refs and which penalized the rates of change of the manipulated inputs least (which have the highest controlled-variable weight to input rate of change weight ratios [10, 10; 140/39, 200, respectively]). However, the worst controlled variable responses were obtained with the guidelines suggested in ref which has the lowest controlled-variable weight to input rate of change weight ratios of 87/49,000 and 58/69,000.…”

“…Burgos et al, 62 and Tsoeu and Koetje 52 ) when the step change was made in the y 2 set point (no model-process mismatch).…”