1338. Model Predictive Control to organize the first vaccination wave during an epidemics (Application to COVID-19 for Wallonia – Belgium)
Invited abstract in session MB-20: Just and ethical sustainability transitions, stream OR and Ethics.
Monday, 10:30-12:00Room: 45 (building: 116)
Authors (first author is the speaker)
| 1. | Morgane Dumont
|
| HECLiege, HECLiege - University of Liege | |
| 2. | Candy Sonveaux
|
| Naxys, University of Namur | |
| 3. | Mirko Fiacchini
|
| Control Systems Department, GIPSA-lab |
Abstract
In case of a new epidemic, decision makers have huge responsibilities in terms of economic impact and public health security. When a vaccine becomes available, they need to strategically define who is going to be vaccinated and when. Moreover, the strategy could need an adaptation after a few days or weeks to be as effective as possible. We propose to use the Model Predictive Control method in order to have a dynamical choice of who is vaccinated and at which moment (with a maximum number of total daily vaccinations). This method starts from the knowledge of the evolution of the disease without any vaccination and proposes a strategy over a fixed prediction horizon by iteratively adjusting the control inputs (vaccination) based on the ongoing system feedback to optimize the performance (minimize the number of dead individuals). To simulate the disease spread, we consider a compartment model including the susceptible, infected, recovered and deaths compartments. The proposed method is illustrated thanks to data about the COVID-19 in Wallonia (Belgium) and the resulting strategy campaign is compared to the strategy really adopted by the government (decreasing order of ages). We show that the two types of campaigns gave similar results in terms of number of deaths and infected, but the Model Predictive Control approach consumes only 63.95% of the vaccines used by the decreasing age strategy. It is thus economically very interesting.
Keywords
- Control Theory
- Decision Support Systems
Status: accepted
Back to the list of papers