ROSHAN (Rescue Oriented Simulation: Handling and Navigating Fires), is a wildfire simulation tool. ROSHAN integrates the principles of cellular automata with reinforcement learning to simulate wildfire dynamics and automatic handling. entral to this simulation tool is its rendering and simulation of fire spread, achieved by integrating data from the CORINE database. The interactive graphical interface of ROSHAN facilitates real-time monitoring and manipulation of fire scenarios. A key component in ROSHAN is the incorporation of a Reinforcement Learning agent, embodied as a drone, which learns to detect and mitigate fires.
You can read everything about the development here.
These modules go under externals:
https://github.com/ocornut/imgui
https://github.com/aiekick/ImGuiFileDialog
https://github.com/nlohmann/json
sudo apt install libpython3.9-dev
https://github.com/pybind/pybind11
cd openstreetmap
npm install express body-parser
npm install --save-dev nodemon
Download Corine CLC+ Backbone - 10 meter (Year 2018)
https://land.copernicus.eu/pan-european/clc-plus/clc-backbone/clc-backbone?tab=download
Move *CLMS_CLCplus_RASTER_2018_010m_eu_03035_V1_1.tif to /CORINE/dataset/
Install GDAL and GDAL C++ headers
sudo apt install libgdal-dev gdal-bin libsdl2-image-dev
Install SDL2 according to:
https://github.com/libsdl-org/SDL/releases/tag/release-2.26.5 https://github.com/libsdl-org/SDL_image/releases/tag/release-2.6.3
conda create --name roshan python=3.9 libffi==3.3
conda activate roshan
pip install torch torchvision
conda install tensorboard
conda install packaging
pip install transformers[torch] onnxruntime bitsandbytes optimum onnx
cd \ROSHAN
mkdir build && cd build
cmake .. && make -j&(nproc)
cd build
./ROSHAN
cd drone_agent/FireSimAgent
python main.py