How to use CVX in matlab code?

Hello!

I am new to cvx, I have no idea how to use it.

I hope you could help me.

I want to optimize the UAV trajectory, I have several variables that I have to calculate before getting the optimal trajectory which are related to the location of the UAV.

I wonder if I should run my program in matlab first with a fixed point of the UAV location, then apply CVX to find the optimal points?

Thanks in advance,

Answer 1

[Torsten]

Use intlinprog.

No need to linearize anything - your equations are linear in the unknowns.

[Safia]

Hi @Torsten ,

this is just an example but in my real case i wil maximize my objective function under a variable(n) and it will be represent in other constraints, the above is just a part. I just mentioned this to get some ideas how to transform a binary value into linear form?

[Torsten]

Show us your real case. In the above, nothing needs to be linearized.

[Torsten]

What are given data in your model ? What is to be determined ?

Constraints (1) and (2) look strange. The first constraint must be wrong since only one u_ij is present without further reference to i or j. In the second constraint, the u_ij should be inside the sum, shouldn't they ?

[Safia]

@Torsten

Z(t), Q,alpha,Ec,E_f,Eb, R ̃_m^((lb) )(t),L : all are known data.

I have also other variables that should be calculated before applying cvx optimizer.

My question is : if u_i,j is a binary value , the problem can't be solved right?

in the first and second constraint ,i multiplied by u_i,j to satisfy the correspondent elements of u_i,j

it means if u_i,j =1 verify condition

else otherwise

[Torsten]

Use intlinprog. As far as I can see, the problem can be solved by a normal linear mixed-integer optimizer.

[Safia]

But I changed non convex constraint to convex in order to use CVX .. But Cvx can't solve MILP problem

[Torsten]

Why is "intlinprog" not an option for you ? Don't you have a license for the optimization toolbox ?

[Safia]

@Torsten yes i do. I already install cvx in matlab

the type of my problem is solved by cvx. But because i don't have any idea how to use it , i'm asking when exactly i call cvx, after doing my all program with given data or how exactly?

[Torsten]

I have no experience with cvx - so I cannot tell you.

I answered because I don't see why you would need a solver different from "intlinprog". And until now, you couldn't show me a part of the model where "intlinprog" would not be sufficient for a solution.

So if you insist on cvx as solver - for whatever reasons - you should open a new question.

Or try a YouTube Video:

https://www.youtube.com/watch?v=1SVU61VsSMc

[Safia]

@Torsten Thank you very much for you effort,

apologize for asking so many questions , i want t know if i will use "intlinprog" ,how can i write my program :

i want to optimize X in order to get maximum sum of u_ij

but X is a variable that appears in equation of R , so i should calculate R at first but in this case with which X should i calculate R. Hope that you understand the question

[Torsten]

I don't see from your model how X influences R. If X is also a solution vector and influences R(t), you will have to use "ga" instead of "intlinprog" since your problem becomes nonlinear.

So define X(1),...,X(T) as solution variables - then you have values in the course of the optimization to calculate R(1),...,R(T) in the "nonlcon" function of "ga".

[Safia]

@Torsten you will understand more by this example : in my case X(t)=qu(t).

at first

because Rn is non convex so :

after this the constraint becomes :