Install Aerodrome

1.Create and activate conda environment

$ conda create -n aerodrome python==3.9
$ conda activate aerodrome

Compile Environment

Aerodrome has been tested on python>=3.9.0,<3.12, but the binary files included in the source code (C++ compiled files under python/simulator) were generated in a python==3.9 environment. If you intend to use a different Python version or your own custom environment, you will need to recompile the binaries and ensure that the Python versions used for compilation and runtime are consistent.

2.Install Pytorch

Running reinforcement learning code requires the use of the torch library; the source code has been tested on torch==2.6.0, but theoretically, it can run on any torch>1.0.0 and compatible CUDA versions. For example, PyTorch 2.6.0 with CUDA 11.8:

$ pip install torch==2.6.0 --index-url https://download.pytorch.org/whl/cu118

3.Install Aerodrome from source

$ git clone https://github.com/CH4ACKO3/Aerodrome.git --recursive
$ cd Aerodrome
$ pip install .

4.Compiling the C++ Code (Optional)

If you intend to use a custom environment or a Python version that differs from the one used to generate the binary files in the source code (python==3.9), you will need to recompile and install the code:

$ cmake -B build
$ cmake --build build
$ pip install .

The platform and Python version used during compilation will be reflected in the binary file names. For example, a file named *.cp39-win_amd64.pyd indicates that it was compiled on a Windows platform with a 64-bit architecture and in a python==3.9 environment.