# Using Robust Optimization for Diet Problem

• 4 replies
• 83 views

• Enthusiast
• 2 replies

GOOD MORNING,

I TRY TO ADAPT THE PROBLEM DIET FOR USE IN F&B BUSINESSES. HOW CAN I DO TO ROBUST OPTIMIZATION? FOR EXAMPLE THE COST PER PERSON REASULTS SHOWS IN ‘PEAK’ AND ‘BASE’.

I HAVE SEEN THE TWO EXAMPLES OF AIMMS BUT NOT HELP ME. I MEAN THE 'POWER SYSTEM EXPANSION RO' AND 'PRODUCTIONPLANNINGRO'.

THANKS

### 4 replies

Userlevel 5
+4

Hi @Kostas Vasilakakis. Could you share your AIMMS project including the changes you made to include robust optimization? Please include the entire AIMMS project folder in a ZIP file, and attach the ZIP file here.

I want to make the variable 'DailyPerPersonFoodCostInFandB(d)', under uncertainty. (Base 3.5 and peak 4.5).

Data 'RO - Vasilakakis 4-2021', in folder data 'cases'.

I have set a 'DailyPerpersonfoodCostinfandBGoal' a goal as a parameter.

And I look for the Variable 'DailyPerpersonfoodCostinfandb(d)' to give results in range.

You can add other corresponding targets afterwards. I also have a lot of constraints. I expect instructions I will move. I am at your disposal for any clarification.

Thanks you very much.

Yours sincerely,

Kostas Vasilakakis

Userlevel 5
+4

Your model contains several errors:

First, in the objective you use a construct like x(d)/p where p is an uncertain parameter. Uncertain parameters cannot be used in a “nonlinear” operator so you cannot use 1/p. One way of resolving this is by introducing a new variable y(d) and change the definition of variable x(d) into y(d) * p. And use y(d) instead of x(d)/p in the definition of the objective.

Second, in the constraint ‘AchieveDailyPerPersonFoodCostInFandB’ you have an equality constraint of the form x + y + z  = p with p an uncertain parameter. This constraint is infeasible as it cannot be satisfied for different values of p; for example the system x + y + z = 1 and x + y + z = 2 has no feasible solution. Only if you change the variables in this constraint into adjustable variables depending on p (see the Language Reference) this is a feasible construction. But in your model you do not use adjustable variables.

Third, in your model you use binary variables but you defined the math program type as LP. This should be MIP (or Automatic).

I will make the changes and I will inform you of any other questions