Here is an example where we know the viability kernel.
Indeed, let's consider two-dimensional dynamics where the orbit of the points on the circle centered at the origin of the coordinate system and with radius $a$ is the same circle; where the orbit of the points with a norm strictly less than $a$ is a spiral tending towards the origin, and where the orbit of the points with a norm strictly greater than $a$ is a spiral which diverges.
This is the case for the solutions of the equation $x'=f(x)$ with
$$
f(x_1,x_2):=(\frac{||x||-a}{||x||}x_1+x_2 ; -x_1+\frac{||x||-a}{||x||}x_2)
$$
Let us consider a square as a constraint set :
$$
K := \{(x_1,x_2)\in R^2\,|\, |x_1|\leq c \text{ et } |x_2|\leq c \}.
$$
Clearly, the set of points from which trajectories originate that remain within $K$ indefinitely is the disk centered at the origin with radius $a$.
This example makes it possible to test the precision of approximations obtained using generic software.
| Dynamics | Constraint | Target | Viability concept |
|
Continuous time 2-dimensional continuous space $f(x_1,x_2):=(\frac{||x||-a}{||x||}x_1+x_2 ; -x_1+\frac{||x||-a}{||x||}x_2) Parameter : $a$ |
$\{(x_1,x_2)\in R^2\,|\, |x_1|\leq c \text{ et } |x_2|\leq c \}$
Parameter : $c$ |
None | Viability kernel |
Below is a comparison, for $a = 0.7$ and $c=1.5$, between the theoretical result (the disk bounded by the red circle) and the approximate result obtained using the ViabLab software on a regular 2001×2001 grid (the green set bounded by the blue line):
We provide the files needed to calculate the aforementioned approximations using the ViabLab software below.
Please consult the training materials: