0
\$\begingroup\$

This is usually easy but there are things I haven't encountered in this one. This is the problem. $$\dot x=\begin{bmatrix}0&1\\-2&-3\end{bmatrix}x+\begin{bmatrix}0\\2\end{bmatrix}u=Ax+Bu\\y=\begin {bmatrix}0&1\end{bmatrix}x=Cx\\x(0)=\begin{bmatrix}1\\0\end{bmatrix}$$

First I'm asked to calculate $$x(t)=e^{At}x(0)$$

An exponential matrix in a control theory course seems a bit too much but I solved that. $$x(t)=\begin{bmatrix}1&1\\-1&-2\end{bmatrix}\begin{bmatrix}e^{-1t}&0\\0&e^{-2t}\end{bmatrix}\begin{bmatrix}2&1\\-1&-1\end{bmatrix}\begin{bmatrix}1\\0\end{bmatrix}$$ The second and final question asks me to find the expression of the output y(t) for any input u(t).

Looking at the solution manual this is all that's written : $$y(t)=\begin{bmatrix}0&1\end{bmatrix}\int_0^t\begin{bmatrix}-1&-1\\-1&-2\end{bmatrix}\begin{bmatrix}e^{-1(t-τ)}&0\\0&e^{-2(t-τ)}\end{bmatrix}\begin{bmatrix}2&1\\-1&-1\end{bmatrix}\begin{bmatrix}0\\2\end{bmatrix}u(τ)dτ$$

That reminds me of convolution but the integral limits are different. Plus the result from the first question is used. Why would we use the exponential? I can't make the connection and understand the second part. Any ideas?

\$\endgroup\$
13
  • \$\begingroup\$ Can Laplace transform be used in the solution? \$\endgroup\$ Commented Aug 30, 2017 at 13:33
  • \$\begingroup\$ @pasabaporaqui Sure, but I would mostly like to understand what he did here. \$\endgroup\$ Commented Aug 30, 2017 at 13:34
  • 1
    \$\begingroup\$ Inverse Laplace of x(0)/(sI-A) is x(0)e^(At), look at any Laplace table \$\endgroup\$ Commented Aug 30, 2017 at 14:21
  • 1
    \$\begingroup\$ A succinct explanation of the transition matrix is given in the Schaum's Outline series on Feedback and Control Systems by DiStefano et all. It is probably a good book to look at if you're going through problems like these. \$\endgroup\$ Commented Aug 30, 2017 at 14:31
  • 1
    \$\begingroup\$ I've got some introductory notes on state space that I give to my students. I don't think I can upload here, so I might try to condense them. Unless someone else answers! \$\endgroup\$
    – Chu
    Commented Aug 30, 2017 at 15:21

1 Answer 1

1
\$\begingroup\$

In Laplace space:

$$ sX(s)-x(0)=AX(s)+BU(s) $$

so:

$$ X(s)=(sI-A)^{-1}BU(s)+(sI-A)^{-1}x(0) $$

the second term, due to the initial condition, produces:

$$ x_0(t)=e^{At}~x(0) $$

while the fist term, due to the incoming signal, produces:

$$ x_u(t)=\int_0^te^{A\tau}~B~u(t-\tau)d\tau=\int_0^te^{A(t-\tau)}~~B~u(\tau)~d\tau $$

(where calculation of \$ e^{At} \$ can be seen here )

Finally:

$$ y(t)=Cx(t)=C(x_u(t)+x_0(t)) $$

\$\endgroup\$
16
  • \$\begingroup\$ Very helpful. I assume I can also find X(s), replace in Y(s)=CX(s) and then taking the inverse Laplace transform find y(t). Is that correct? \$\endgroup\$ Commented Aug 30, 2017 at 19:39
  • \$\begingroup\$ Taken into account that C is a constant, y(t)=Cx(t) and Y(s)=CX(s), you can do the multiplication in time or Laplace space. \$\endgroup\$ Commented Aug 30, 2017 at 19:42
  • \$\begingroup\$ Is there a chance the initial conditions are not in y(t)? I mean C could cancel them out right? \$\endgroup\$ Commented Aug 30, 2017 at 19:53
  • \$\begingroup\$ No, initial conditions can not be "cancelled" (if you are moving at 100 km/h, you can accelerate or decelerate, but not ignore you are moving). If C is not a constant but c(t) it means that the system is not time invariant. In this case, usually is easier calculation in time space y(t)=c(t)x(t) than calculation in Laplace space L-1{convolution(L{c(t)},L{x(t)})} \$\endgroup\$ Commented Aug 31, 2017 at 8:29
  • 2
    \$\begingroup\$ No, do not start by reading books nor taken notes (or going to magisterial class): relax, take time, and thing about what you have in front. Try to visualize and give sense to each part of the problem. Books are for reference, or to stablish an ordenation in the matter. \$\endgroup\$ Commented Aug 31, 2017 at 11:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.