saugat76 / uav_nw_madql_connectivity_tensorflow

Allocation of the resource blocks to the ground user for distributed UAV based cellular system with only the transfer of reward values. Additional improvement over the other setup by the use of Deep Q Learning approach.

UAV_SubBand_Allocation_DQN

Allocation of the resource blocks to the ground user for distributed UAV based cellular system with only the transfer of reward values. Additional improvement over the other setup by the use of Deep Q Learning approach.

Dependencies installation and steps to run on a Conda Environment

1) Installation of Anaconda / Ommit this step if already installed

Install annconda on your machine. For further information follow this link https://docs.conda.io/projects/conda/en/latest/user-guide/install/index.html

2) Open anaconda prompt and create a new conda environment with python 3.10. After that activate the environment

conda activate uavenv

3) Install the following dependencies

Install gym, tensorflow-gpu with it's required dependencies

conda install gym
conda install pip
pip install tensorflow-gpu
conda install matplotlib
conda install -c anaconda cudatoolkit

4) Open your IDE through the conda environment

Preferred Method

If you already have Visual Studio Code. You can enter the following command in the same conda enviroment to open up the IDE.

code

Alternative Method

You can also open up your own IDE and change the interpreter setting to the python.exe file with the .conda/env/uavenv folder before running the code

5) Optional step to clone git

Inside visual studio code click on "New File" and in file and inside explorer tab click on "Clone Repository". Enter the URL of this git repo to run this code. https://github.com/saugat76/UAV_SubBand_Allocation_DQN.

6) Adjust the memory limit

Adjust the memory limit and the configuration of GPU according to you system capability.

NOTE : These code use tensorflow and CPU for the training process so is comparatively slow compared with the newer deployement using pytorch library.