Académique Documents
Professionnel Documents
Culture Documents
Course structure
Three parts 1. SISO control revisited 2. Multivariable control a. Multivariable linear systems b. Multivariable loop shaping (today) c. Linear quadratic control d. H2, H1 and design example e. Decentralized control and decoupling f. Glover-McFarlane loop shaping 3. Systems with hard constraints
continued this lecture Recap of last lecture Fundamental limitations and robustness of MIMO systems Modern multivariable loop shaping
Note: should push up smallest singular value at low frequencies, and make sure that largest singular value is small at high frequencies.
where
Note: the infinity norm computes the maximum amplifications across frequency (sup) and input directions ( )
Definition. The zero polynomial of G(s) is the greatest common divisor of the maximal minors of G(s), normed so that they have the pole polynomial of G(s) as denominator. Theorem. The zeros of the G(s) are the roots of its zero polynomial.
As for scalar systems, a zero at z implies that there exists a signal on the form u(t)=vze-zt for t 0, and u(t)=0 for t<0, along with initial values x(0)=xz so that y(t)=0 for t 0
EL2520 Control Theory and Practice Elling W Jacobsen jacobsen@kth.se
2
(u1, u2)=(1, 0)
1 0 -1 0 2 2 4 6 8 10
(u1, u2)=(0, 1)
1 0 -1 0 2 2 4 6 8 10
(u , u )=(1, -1)
1 0 -1 0 2 4 6 8 10
EL2520 Control Theory and Practice
is that
Note. Implies bandwidth limitations on maximum singular value, but not on the individual channels (more in the example at the end of this class)!
singular if (1,2) element is perturbed from -86.4-88.9. However, reasonable perturbations are on the form
with
Note: one block per uncertain perturbation, e.g. distillation column example has a single uncertain parameter.
EL2520 Control Theory and Practice Elling W Jacobsen jacobsen@kth.se
The structured singular value: find smallest structured perturbation (in terms of ) that renders closed-loop unstable, i.e. causes
Then, the structured singular value Powerful framework for analysis and design of robust MIMO controllers Details out of the scope of this course Learn more in our graduate (PhD-level) course on robust control!
Forbidden area
Assumption:
H1 optimal control
Finding a controller that satisfies
Ex. 3. Start shaping most important transfer matrix, then add one by one
Ex. 4. When channels are very different, use diagonal weights
Example
Consider the system with RHP zero
A first design
Since system has RHP zero at z=0.5, a reasonable weight is
The mixed sensitivity design achieves min=2.79. Reasonable sensitivities, but poor time-domain performance.
A second design
Can shift bandwidth limitation from one channel to the other (i.e. alter singular vectors; limitation on maximum singular value remains)
Time response on channel two much better, constraint on bandwidth of maximum singular value of S still present.
Conclusions
More elements of multivariable system theory Multivariable frequency responses, gains, poles/zeros, directions
Key concepts from SISO analysis carry over fundamental limitations, robustness, loop shaping Modern loop shaping Mixed sensitivity design: minimizing H1 norm of extended system Optimal solution is state-feedback plus observer Tuning knobs for design are weight functions Weight selection: part art, part science (must practice!)