This question is generated from Example 6.3 in Slotine and Li "Applied Nonlinear Control", Prentice Hall, 1991

A nonlinear system

$$ \begin{bmatrix}\dot{x_1}\\\dot{x_2}\end{bmatrix}=\begin{bmatrix}{x_2}^3+u\\u\end{bmatrix}$$ $$ y=x_1 $$ as the control objective to make $y$ track $y_d$, and the tracking error is $e=y-y_d$.

It is, $\dot{y}=\dot{x_1}$ and choosing the control law

$$ u=-{x_2}^3-e+\dot{y_d} $$ gives $$ \dot{e}+e=0 $$ which is stable and converging to zero as the time $t\rightarrow\infty$ (or in other words, the error equation has one pole in -1).

The same control input $u$ applies also to the second equation of the nonlinear system, representing the internal dynamics. With the choice of $u$ as above, $\dot{x_2}=u$ yields $$ \dot{x_2}+{x_2}^3=\dot{y_d}-e. $$

With the choice of a bounded $\dot{y_d}$ and $e$ (bounded since $\dot{e}+e=0$), then it is $$ |\dot{y_d}-e|\leq D, $$ being $D$ a positive constant.

From now on my question starts, as I would like to know what are the math steps to derive the following conclusion:

the example concludes that $|x_2|\leq D^{1/3}$ (i.e. $x_2$ is bounded too!), since $\dot{x_2}<0$ when $x_2>D^{1/3}$, and $\dot{x_2}>0$ when $x_2<D^{1/3}$. Can someone explain how to derive these inequalities?

The example demonstrates that the chosen control law for $u$, i.e. $u=-{x_2}^3-e+\dot{y_d}$ -chosen from the first state equation- does not causes the second state $x_2$ to become unbounded.


Let us rewrite your equation as $\dot{x}_2 = -x_2^3 + d(t)$, where $|d(t)|\le D$ for all $t$. If $x_2$ is positive, then $\dot{x}_2$ is negative for all $x_2>D^{1/3}$. If $x_2$ is negative, then $\dot{x}_2$ is positive for all $x_2<-D^{1/3}$. It implies that the set $|x_2^3|\le D$ is invariant and attractive: if the initial condition $x_2(t_0)$ belongs to the set, then the trajectory reamins in the set. If the initial condition $x_2(t_0)$ is outside the set, then it will converge to the set.

Thus, $x_2(t)$ is bounded. However, the claim that $|x_2|\le D^{1/3}$ is valid only if you know for sure that the initial condition satisfies this inequality, or if you consider it as $t \to \infty$.

  • $\begingroup$ Thank you @Arastas, you made it very clear. So, if I understand well, we are trying to ensure that $x_2$ is always bounded. So we state the condition for $\dot{x_2}>0$ and $\dot{x_2}<0$ (as I assume we don't know $\dot{x_2}$ itself, but we try to estimate what happens in each case)... overall this approach is looking at what happens to $x_2$ when the control law chosen based on controlling $x_1$ is applied $\endgroup$ – Lello Nov 17 at 22:11
  • 1
    $\begingroup$ Yes, you are right. There are also more general ways to study it, i.e., the Lyapunov functions. But for the scalar case is more simple. $\endgroup$ – Arastas Nov 18 at 9:17

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.