Jump to content

LMIs in Control/pages/Switched systems H2 Optimization

From Wikibooks, open books for an open world

LMIs in Control/pages/Switched systems H2 Optimization


Switched Systems Optimization

The Optimization Problem

[edit | edit source]

This Optimization problem involves the use of the State-feedback plant design, with the difference of optimizing the system with a system matrix which "switches" in properties during optimization. This is similar to considering the system with variable uncertainty; an example of this would be polytopic uncertainty in the matrix . which ever matrix is switching states, there must be an optimization for both cases using the same variables for both.

This is first done by defining the 9-matrix plant as such: , , , , , , , , and . Using this type of optimization allows for stacking different LMI matrix states in order to achieve the controller synthesis for .

The Data

[edit | edit source]

The data is dependent on the type the state-space representation of the 9-matrix plant; therefore the following must be known for this LMI to be calculated: , , , , , , , , and . What must also be considered is which matrix or matrices will be "switched" during optimization. This can be denoted as .

The LMI: Switched Systems Optimization

[edit | edit source]

There exists a scalar , along with the matrices , , and where:


Where is the controller matrix. This also assumes that the only switching matrix is ; however, other matrices can be switched in states in order for more robustness in the controller.


Conclusion:

[edit | edit source]

The results from this LMI gives a controller gain that is an optimization of for a switched system optimization.


Implementation

[edit | edit source]
% Switched System H2 example
% -- EXAMPLE --

clear; clc; close all;

%Given
A  = [ 1  1  0  1  0  1;
      -1 -1 -1  0  0  1;
       1  0  1 -1  1  1;
      -1  1 -1 -1  0  0;
      -1 -1  1  1  1 -1;
       0 -1  0  0 -1 -1];
  
B1 = [ 0 -1 -1;
       0  0  0;
      -1  1  1;
      -1  0  0;
       0  0  1;
      -1  1  1];

B21= [ 0  0  0;
      -1  0  1;
      -1  1  0;
       1 -1  0;
      -1  0 -1;
       0  1  1];

B22= [ 0   0   0;
      -1   0   1;
      -1   1   0;
       1   1   0;
       1   0   1;
       0  -3  -1];

C1 = [ 0  1  0 -1 -1 -1;
       0  0  0 -1  0  0;
       1  0  0  0 -1  0];

D12= [ 1    1    1;
       0    0    0;
       0.1  0.2  0.4];

D11= [ 1  2  3;
       0  0  0;
       0  0  0];
   
%Error
eta = 1E-4;

%sizes of matrices
numa  = size(A,1);    %states
numb2 = size(B21,2);  %actuators
numb1 = size(B1,2);   %external inputs
numc1 = size(C1,1);   %regulated outputs

%variables
gam = sdpvar(1);
Y   = sdpvar(numa);
Z   = sdpvar(numb2,numa,'full');
W   = sdpvar(numc1);

%Matrix for LMI optimization
M11 = Y*A'+A*Y+B21*Z+Z'*B21'+B1*B1';
M12 = Y*A'+A*Y+B22*Z+Z'*B22'+B1*B1';
M2  = [Y            (C1*Y+D12*Z)'  ;
       C1*Y+D12*Z   W              ];

%Constraints
Fc = (M11 <= 0);
Fc = [Fc; M12 <= 0];
Fc = [Fc; trace(W) <= gam];
Fc = [Fc; M2 >= zeros(numa+numc1)];

opt = sdpsettings('solver','sedumi');

%Objective function
obj = gam;

%Optimizing given constraints
optimize(Fc,obj,opt);

%Displays output Hinf gain
fprintf('\n\nHinf for H2 optimal state-feedback problem is: ')
display(value(gam))

F = value(Z)*inv(value(Y)); %#ok<MINV>

fprintf('\n\nState-Feedback controller F matrix')
display(F)
[edit | edit source]


Return to Main Page:

[edit | edit source]