2944. A MIP with warm start to solve the IHTC problem
Invited abstract in session MC-10: Integrated Healthcare Timetabling Competition II, stream Automated Timetabling.
Monday, 12:30-14:00Room: Clarendon SR 1.06
Authors (first author is the speaker)
| 1. | Lotte Berghman
|
| ORTEC | |
| 2. | Per Kampman
|
| 3. | Wouter Kool
|
| ORTEC | |
| 4. | Judith Mourits
|
| ORTEC | |
| 5. | Martijn van Brink
|
| ORTEC | |
| 6. | Eva van Rooijen
|
| ORTEC BV | |
| 7. | Qing Chuan Ye
|
| ORTEC B.V. |
Abstract
We propose a solution method for the IHTC problem (ihtc2024_problem_specification.pdf) that employs a combination of a sequential construction algorithm and a mixed integer program (MIP).
A patient construction sequentially tries to admit a patient and assign it to a room and operating theater, adhering strictly to the feasibility constraints. This is repeated until no more patients can be admitted. The algorithm is randomized by randomly skipping candidate patients during constructions. Afterwards, nurse constructions loop over the days and rooms and simply assigns the nurse that will result in the least increase in cost.
Our MIP introduces additional variables that represent combinations of decisions (i.e. a patient treated by a nurse on a specific day in a specific room). When the MIP is too large to solve (large instances), we sequentially solve different subMIPs. In the first stage, we only assign patients to days and operating theaters. This solution is warm started with the patient and day assignments of the best construction. In the second stage, we fix days and operating theaters and assign patients to rooms. Finally, we fix the day/operating theater/room assignments and solve the nurse assignment. The best overall result is returned.
Keywords
- Timetabling
- Rostering
- Health Care
Status: accepted
Back to the list of papers