Table 5:
The best-found solutions for the integrated planning and scheduling case studies using DOMINO with various data-driven optimization solvers. The scheduling levels for Example 1 and 2 are solved with 5 event points whereas for Example 3, 6 event points are used. Standard deviation is calculated with the scaled objective function values reported below. The bold values reflect the best-performing solver among the tested.
| NOMAD | COBYLA | ARGONAUT | EGO | ISRES | |
|---|---|---|---|---|---|
| Example 1 | |||||
|
| |||||
| LP-MILP/1000 | 7.11 | 7.42 | 7.45 | 8.16 | - |
| Infeasible Runs (Out of 10) | 0 | 1 | 0 | 2 | 10 |
| Std. Dev. | 0.33 | 0.78* | 0.28 | 0.73* | - |
| Tot. No. Samples | 1508 | 187 | 3028 | 97 | - |
| NLP-MIQP/1000 | 6.20 | 6.41 | 6.29 | 6.46 | 6.66 |
| Infeasible Runs | 0 | 1 | 0 | 2 | 9 |
| Std. Dev. | 0.04 | 0.26* | 0.06 | 0.2 | 0* |
| Tot. No. Samples | 658 | 131 | 4258 | 106 | 8507 |
|
| |||||
| Example 2 | |||||
|
| |||||
| LP-MILP/1000 | 13.52 | 15.03 | 14.49 | 15.32 | - |
| Infeasible Runs | 0 | 9 | 0 | 3 | 10 |
| Std. Dev. | 0.28 | 0* | 0.65 | 0.9* | - |
| Tot. No. Samples | 2972 | 396 | 2912 | 210 | - |
| NLP-MIQP/1000 | 12.21 | - | 12.33 | 12.55 | - |
| Infeasible Runs | 0 | 10 | 0 | 3 | 10 |
| Std. Dev. | 0.09 | - | 0.1 | 0.23 | - |
| Tot. No. Samples | 2527 | - | 2358 | 127 | - |
|
| |||||
| Example 3 | |||||
|
| |||||
| LP-MILP/1000 | 102.08 | 109.46 | 112.73 | 120.33 | 130.20 |
| Infeasible Runs | 0 | 3 | 0 | 0 | 4 |
| Std. Dev. | 0.92 | 10.66* | 3.4 | 3.21 | 4.26* |
| Tot. No. Samples | 9572 | 1015 | 1165 | 376 | 7950 |
| NLP-MIQP/1000 | 224.69 | 224.94 | 233.12 | 241.15 | 277.39 |
| Infeasible Runs | 0 | 4 | 0 | 0 | 7 |
| Std. Dev. | 0.2 | 27.65* | 5.44 | 7.84 | 8.24* |
| Tot. No. Samples | 9426 | 4097 | 1169 | 288 | 6409 |
*
Standard deviation calculation excludes infeasible runs.