In optimization problem, I have two constraints for decision variables $x_1$ and $x_2$ as follow

$0 \leq x_1 \leq u_1$ and $0 \leq x_2 \leq u_2$

I was wondering how can I replace these 2 constraints with just one constraint in terms of $x_1,x_2,u_1,u_2$. Hence, the goal is to reduce the number of constraints. I dont care if this results in nonlinear constraints.

  • $\begingroup$ I dont´t see a way to transform the two constraints into one constraint. Even it could be made formally you would still have technically two constraints. $\endgroup$ Sep 2 '17 at 18:39
  • $\begingroup$ May I ask what motive you have here? The answer isn't going to be any better practically. $\endgroup$ Sep 3 '17 at 2:28
  • $\begingroup$ It seems like an 'ill posed' problem. How do you count a "constraint"? Because what you wrote, in as I see it, is actually four constraints. So if I interpret the term 'constraint', I can write it as one constraint: $(x_1, x_2) \in C$, where $C = [0, u_1] \times [0, u_2]$. $\endgroup$
    – Alex Shtof
    Sep 3 '17 at 3:21

Just replace the constraints with this one:

$$\max\{-x_1,x_1-u_1,-x_2,x_2-u_2\}\leq 0.$$

More generally, the inequality constraints

$$g_i(x)\leq 0,\;i=1,\ldots,m$$ can be replaced by the single constraint

$$\gamma(x)=\max_{i=1,...,m}\{g_i(x)\}\leq 0.$$ If $g_i$ is convex for each $i,$ then $\gamma $ is also convex. Hope this helps.


If one of the $u_i<0$, there is no solution.

If one of the $u_i=0$, we know $x_i=0$ and you can focus on another constraint.

Assuming $u_1$ and $u_2$ are both positive,

$$0 \leq \frac{x_i}{u_i} \leq 1$$

$$-\frac12 \leq \frac{x_i}{u_i} - \frac12 \leq \frac12$$

$$\left|\frac{x_i}{u_i}-0.5 \right| \leq 0.5$$

$$\max_{i=1,2} \left|\frac{x_i}{u_i}-0.5\right| \leq 0.5$$


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.