Controllability of state space equation

[dabargo]

Homework Statement

I have calculated the following mechanical system

<br /> <br /> \left( \begin{array}{c}\dot{\mathbf{x_1}}(t) &amp; \dot{\mathbf{x_2}}(t) \end{array} \right) = \left( \begin{array}{cc}-0.5 &amp; 0 &amp; 0 &amp; -1\end{array} \right) \cdot \left( \begin{array}{c}x_{1}(t) &amp; x_{2}(t)\end{array} \right) + \left( \begin{array}{c}0.5 &amp; 1\end{array} \right)\cdot u(t)<br /> <br />

<br /> <br /> \left( \begin{array}{c}y_{1}(t) &amp; y_{2}(t)\end{array} \right) = \left( \begin{array}{cc}1 &amp; 0\end{array} \right) \cdot \left( \begin{array}{c}x_{1}(t) &amp; x_{2}(t)\end{array} \right) <br /> <br />

The question is to find the expression for u(t) that brings the system to its restposition in 2 seconds. Afterwards i have to simulate this time response in MATLAB with the function lsim.

The initial conditions of the system are:
<br /> <br /> x_1(0) = 10<br /> <br />
<br /> <br /> x_2(0) = -1<br /> <br />

Homework Equations

<br /> <br /> u(t) = -B^T\exp^{A^T(t_1-t)}W_c^{-1}(t_1)[\exp^{At_1}x_0-x_1]<br /> <br />
<br /> <br /> W_c(t_1) = \int_0^{t_1} \exp^{A\gamma}BB^T\exp^{A^T\gamma} d\gamma<br /> <br />

The Attempt at a Solution

I calculated u(t) with MATLAB with the following code

A = [-0.5 0;0 1];
B = [0.5; 1];
C = [1 0];
D = 0;
syms t;
Wc = int(expm(A*t)*B*transpose(B)*expm(transpose(A)*t),0,2);
u = -transpose(B)*expm(transpose(A)*(2-t))*inv(Wc)*(expm(A*2)*[10;-1]-[0;0]);
sys = ss(A,B,C,D);
t = 0:0.1:2;
u = subs(u,t);
lsim(sys,u,t)

But the result i get with lsim doesn't fulfill my expectations at all.
The result i get is given in the lsimresults.bmp.
While i expect it to curve a bit down/up the original time responses (1st and 2nd thumbnails) where u(t) = 0, so that the amplitude of the system becomes 0 at 2 seconds.

So, if you there is anybody who can give me some help where it goes wrong. Or if i interpretate something wrong maybe?

 

[Eidos]

Homework Equations

<br /> <br /> u(t) = -B^T\exp^{A^T(t_1-t)}W_c^{-1}(t_1)[\exp^{At_1}x_0-x_1]<br /> <br />
<br /> <br /> W_c(t_1) = \int_0^{t_1} \exp^{A\gamma}BB^T\exp^{A^T\gamma} d\gamma<br /> <br />

Where did you get those equations, your system is not doing what you want it to do I agree.

I've tried using state feedback to solve your problem, here is what I got:

u=-Kx

where x=[x_1 \, x_2]^{T}
and K=[K_1 \, K_2]

You then choose the eigenvalues of the closed loop system
(sI-A+BK) by varying K such that they have the properties you want.

In this case you want a settling time of 2s, so I would make both of my eigenvalues equal to -2, which gives us the desired characteristic polynomial of

(s+2)(s+2)=s^2+4s+4

Now to go about solving for K,

the controllers characteristic polynomial is
det(sI-A+BK)=\alpha_2s^2+\alpha_1s+\alpha_0

All you do now is compare the desired characteristic polynomial with the actual characteristic polynomial and solve for K_1, K_2 such that the co-efficients of the two characteristic polynomials are equal i.e.

\alpha_2 = 1
\alpha_1 = 2
\alpha_0 = 4

For this system there is a pole that no matter what feedback you give the system, will not move higher

the pole is located at approx. -0.74

If the settling time we are talking about is 0-63% of final value then you are ok.
The other pole you can shift around at whim using this technique.

The gains I used are K=[1 \, 2] to give poles at -0.73 and -5.76

I don't have access to MATLAB now but try its "place" command
maybe you can have better luck that way