How to change the RHS for a set of constraints and solve the problem iteratively using JuMP?

I have an optimization problem where i want to change the RHS of a set of constraints and solve the model each time.

The RHS Demand_specific is the input to the model. I want to test the model for different fluctuations in demand values, so i am changing the demand matrix iteratively.

using JuMP, Gurobi 

for t in 1:n_echelons
    for g in 1:n_customers
         for k in 1:n_customers
             for p in 1:n_products
                 @constraint(IMTLP, Demandsconst, Direct_shipping[k,g,p,t]+sum(Endhaul_shipping[j,g,k,p,t] for j in 1:n_hubs)>=Demand_specific[g,k,p,t])

OptimalCosts = zeros(n_instances)

for i in 1:n_instances
    JuMP.setRHS(Demandsconst, (2i/n_instances)*Demand_specific)
    status = solve(IMTLP)
    z = getobjectivevalue(IMTLP)
    OptimalCosts[i] = z

I am getting the following warning:

Warning: A variable or constraint named Demandsconst is already attached to this model. If creating constraints programmatically, use the anonymous constraint syntax con = @constraint(m, …).

Please try to format your code to make it more readable. This is explained in PSA: make it easier to help you.

It looks like you are create many constraints in nested loops, but always reuse the same identifier Demandsconst. You can not refer to all of these constraints with a single name.

Try moving the loops “into” the @constraint macro call and storing the constraint references in an array, like @constraint(IMTLP, Demandsconst[t in 1:n_echelons, ...], ...) as described in the JuMP docs.

Later you will have to update the right-hand side of all these constraints individually, I guess.

1 Like

Maybe it would also make sense to introduce a dummy variable that you place as the right-hand side. Then you could fix that variable to different values and would not have to touch the constraints.

i can’t use the first approach due to my problem structure, i will give the second method a try.


You could just use anonymous constraints,, after which you can store the generated constraint references in any container you like.

Thanks it did work, i shouldn’t have used nested loops,

I did it like, and then i can use JuMP.setRHS:

@constraint(model, con[i = 1:2, j = 2:3], i * x <= j + 1)

Thanks, the documentation made it clear to me how the @constraint macro works