Conceptualize multi-agent interactions.

TheVeggydude / MARL-lockdown

A Multi-Agent Reinforcement Learning project that learns to solve the Coordinated Lockdown problem.

MIT License

3 stars 0 forks source link

Conceptualize multi-agent interactions. #2

Closed TheVeggydude closed 4 years ago

TheVeggydude commented 4 years ago

Right now there is only a single agent going through the infection cycle. This needs to be extended to a multitude of agents. This step is for figuring out the theoretical underpinnings of the connections between agents.

Perhaps they could be done by having a massive matrix containing the connections between each agent.
Perhaps the agent compartment values could even be in a matrix?
The parameters could be in another matrix, so maybe updating could be done through matrix multiplications.

TheVeggydude commented 4 years ago

In the initial designs, have a flat transfer rate, e.g. country A will always have 10% of its population (equal across all compartments) move to country B.

Maybe later on we can look into modelling the behavior in a more realistic manner, such as infected people travelling less due to e.g. hospitalization.