Multi objective result changes from single objective, when model is exact the same

回答済み

Timo Kaiser

2025年04月10日 06:46 編集済み

• 
• 

Hello All, 

How can it be that my optimization result changes when I change my single objective model to a multi-objective model with only one objective, where everything remains the same exact for objective definition through model.setObjectiveN() instead of model.setObjective()

E.g.:

Changing this:

model.setObjective(
            quicksum(
                itinerary_fam_df.loc[i, "Fare"] * x[n, i] 
                for n in O_D.index
                for i in usable_its_if[(f"{n}")]
                if (n, i) in x
            )
            - quicksum(
                aircraft_data.loc[k, "c_inv(EUR)"] * z[k, r] 
                for k in aircraft_data.index for r in max_aircraft_per_type_range
            )
            - quicksum(
                (
               w[(iteration - 1)][scenario][f'z[{k},{r}]'] 
                 - rho[iteration] * x_avg[(iteration - 1)][f'z[{k},{r}]']
                 + rho[iteration]/2 
                )* z[k, r]
                for k in aircraft_data.index
                for r in max_aircraft_per_type_range
            )
            - quicksum(
                (
                    eta_leg_fam.iat[l, k] * Leg_fam.loc[l, "Distance"] * p_fuel
                    + Leg_fam.loc[l, f'{aircraft_data.loc[k, "AC Code"]} cost']
                )
                * f[l, k]
                for l in Leg_fam.index
                for k in aircraft_data.index
                if not np.isnan(eta_leg_fam.iat[l, k])
                if not np.isnan(Leg_fam.loc[l, f'{aircraft_data.loc[k, "AC Code"]} cost'])
                if (l, k) in f
            )
 
        )

to this

  model.setObjectiveN(
            quicksum(
                itinerary_fam_df.loc[i, "Fare"] * x[n, i] 
                for n in O_D.index
                for i in usable_its_if[(f"{n}")]
                if (n, i) in x
            )
            - quicksum(
                aircraft_data.loc[k, "c_inv(EUR)"] * z[k, r] 
                for k in aircraft_data.index for r in max_aircraft_per_type_range
            )
            - quicksum(
                (
               w[(iteration - 1)][scenario][f'z[{k},{r}]'] 
                 - rho[iteration] * x_avg[(iteration - 1)][f'z[{k},{r}]']
                 + rho[iteration]/2 
                )* z[k, r]
                for k in aircraft_data.index
                for r in max_aircraft_per_type_range
            )
            - quicksum(
                (
                    eta_leg_fam.iat[l, k] * Leg_fam.loc[l, "Distance"] * p_fuel
                    + Leg_fam.loc[l, f'{aircraft_data.loc[k, "AC Code"]} cost']
                )
                * f[l, k]
                for l in Leg_fam.index
                for k in aircraft_data.index
                if not np.isnan(eta_leg_fam.iat[l, k])
                if not np.isnan(Leg_fam.loc[l, f'{aircraft_data.loc[k, "AC Code"]} cost'])
                if (l, k) in f
            ),
            index=0,
            priority=1,
            weight=1
            #reltol= obj_tolerance
        )

changes the Obj value from 1.7e06 to 2.6e06. How can that be?

 

関連

• Multiple Objectives

1

• 

  Riley Clement

  • Gurobi Staff

  2025年04月10日 04:39

  Hi Timo,

  How can that be?

  It shouldn't.  Can you write your two models to MPS files and upload them to a file sharing site (such as DropBox, Google Drive, public Github repo etc) so we can take a look?

  - Riley

  0
• 

  Timo Kaiser

  2025年04月10日 06:19

  • 
  • 

  Hi Riley, 

   

  Thanks for looking into it. I saw you also commented the more detailed forum post on it. I will edit the post to have the indentation, sorry about this. 

   

  Here are the model files: 

  https://tumde-my.sharepoint.com/:f:/g/personal/timo_kaiser_tum_de/Er7GLkp9K-VPrLSt3UnRJgsBzmmWp1C3ckXaKxaWbuzMHQ?e=qUXTk7

   

  Best 

  Timo

  0
• 

  Riley Clement

  • Gurobi Staff

  2025年04月10日 11:28

  Hi Timo,

  Which version of Gurobi are you using?  There was a bug in v12.0.0 related to setting objectives in multiobjective models - see "Incorrect handling of duplicate indices in setObjectiveN" from Known issues in Gurobi 12.0.0

  This bug was fixed in v12.0.1 so if you're not using that then please test it.

  - Riley

  0
• 

  Timo Kaiser

  2025年04月10日 13:00

  • 
  • 

  Hi, 

  That seems to have been the problem. Looks like it's fixed now. 

   

  Thank you!

   

   

  1

サインインしてコメントを残してください。