Skip to content

Controlling forest fires in a simulation using reinforcement learning, as part of a bachelors project at the University of Groningen. This is a Python implementation of the original Wildfire-Control project.

License

Notifications You must be signed in to change notification settings

dashdeckers/Wildfire-Control-Python

Repository files navigation

Wildfire-Control-Python

The custom environment simulates the spread of fire from a 2D, birds-eye-view perspective. To contain the fire, the agent (a bulldozer) should learn to dig a road around the fire, enclosing it completely. By doing so, we take away the fuel that the fire needs to spread further.
Follow these instructions with python 3.6 in a virtual environment!

Install dependencies:

pip install -r requirements.txt

Create A* files:

make -C pyastar/

Let the algorithm learn and then let it play:

python main.py -r -m {amount_of_memories} -e {amount_of_episodes} -t {DQN/SARSA/DDQN/BOTH} -n {name}

Play as the agent:

python main.py -t Human

For more information:

python main.py -h

About

Controlling forest fires in a simulation using reinforcement learning, as part of a bachelors project at the University of Groningen. This is a Python implementation of the original Wildfire-Control project.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •