Jeevesh8 / FireFighter-RL

Implementation of some Reinforcement Learning algorithms for the Fire Fighter problem on graphs

FireFighter-RL

Implementation of some Reinforcement Learning algorithms for the Fire Fighter problem on graphs

An Example Of Learned strategy

The model is allowed to only defend two vertices at a timestep. Blue vertices denote defended ones, yellow the burning ones and purple ones are neither burning, nor defended.

With burn probability 0.5

The image below shows the decisions taken by the learned model in the various states.

With burn probability 1

Different decisions are taken by the learned agent, it decides to let the vertex in the burning row, burn, as there is 100% chance that 2 vertices will be burnt, unlike before.