1
$\begingroup$

I have a DH matrix (Denavit-Hartenberg) of a two link manipulator having differential equation of the form:

$$ \begin{bmatrix} \tau_1 \\ \tau_2 \end{bmatrix}= \begin{bmatrix} k_1+k_2\cos\theta_2 & k_3+k_4\cos\theta_2 \\ k_5+k_6\cos\theta_2 & k_7 \end{bmatrix} \begin{bmatrix} \ddot{\theta}_1 \\ \ddot{\theta}_2 \end{bmatrix}+ \begin{bmatrix} k_8\dot{\theta}_2^2 - k_9\dot{\theta}_1\dot{\theta}_2 \\k_{10}\dot{\theta}_1^2 \end{bmatrix}+f(\theta_1,\theta_2) $$

where $f$ is a nonlinear function. $k_i$ are constants

How can such a nonlinear coupled differential equation be expressed in state space form? How to decouple terms which contain product of both states?

Edit : I want to design a state observer for this, so wanted it to be represented in state space form.

$\endgroup$
  • $\begingroup$ You cannot state the linear state space representation without linearizing the system. $\endgroup$ – MrYouMath Nov 6 '17 at 9:21
  • $\begingroup$ I am not having the main problem in linearizing it. Let's assume this question is linearized about some point. How to decouple the terms where the two states are multiplied? $\endgroup$ – Aniket Sharma Nov 6 '17 at 9:23
  • $\begingroup$ The linearization automatically eliminates all terms that are nonlinear e.g. $ \dot{\theta}_1\dot{\theta}_2 $. In your question, you asked how to state the state space representation and this is what @LutzL has answered. What you want is to get the linear time-invariant form of the system $\endgroup$ – MrYouMath Nov 6 '17 at 9:33
1
$\begingroup$

You have an equation of the form $$ r=M(θ)\ddot θ+b(θ,\dot θ) $$ Introduce $\omega = \dot θ$ to find the first order system $$ \begin{bmatrix}\dot θ\\\dot ω\end{bmatrix} = \begin{bmatrix}ω\\M(θ)^{-1}(r-b(θ,ω))\end{bmatrix} $$ provided the matrix $M(θ)$ is invertible.

$\endgroup$
  • $\begingroup$ But the new matrix that you formed still has theta and omega coupled in the function b. How to express this in dX/dt = AX+BU form? $\endgroup$ – Aniket Sharma Nov 6 '17 at 9:03
  • $\begingroup$ You can't as the system is deeply non-linear. If you have the goal to construct a controller you should post it as relevant information in your question. $\endgroup$ – LutzL Nov 6 '17 at 10:06
  • $\begingroup$ I want to design a state observer and controller. So I want to represent it in state space form. $\endgroup$ – Aniket Sharma Nov 6 '17 at 10:11
  • $\begingroup$ @AniketSharma The first order differential equation in this answer is a state space form. Only it is nonlinear, so you would have to linearize it in order to obtain the form $\dot{x}=A\,x+B\,u$. $\endgroup$ – Kwin van der Veen Nov 6 '17 at 15:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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