Control of Covid-19 using Agent-based modelling with Reinforcement learning.

Polymenidis, Stergios ; Wawrzyniak, Bartosz

4. term

Control and Automation, Master

2021

2021-06-03

96 pages

Humanity has been struggling with infectious diseases since the dawn of time. These diseases pose a great threat to society and the fight against them is sometimes challenging. Currently, the world is struggling with a pandemic of a virus that causes a disease called Covid-19. Disease is rarely fatal but spreads quickly and becomes easily out of control. Controlling the spread of the virus has become a challenge for governments. In this project, we propose to use a reinforcement learning algorithm to find the optimal policy to keep hospitalized and severe (requiring a respirator) cases within the imposed thresholds. We use an agent-based modelling technique to simulate society and the spread of the virus within it. Following the actions of governments regarding the pandemics, we have established a list of policies that, cause reactions similar to those in the real world. We apply the model-free value iteration reinforcement learning algorithm to the model to find a sequence of policies that will allow to control the spread of the disease and keep hospitalized and those requiring a respirator at a level that will not overload health care. We create three models with different complexities to test the operation of the algorithm. We simulate two models and the results show that the algorithm can find the desired sequence of policies.

Covid-19 ; reinforcement learning ; markov decision process ; agent-based learninig

Download
View record in AAU Student Projects