This repository provides an implementation of our deep reinforcement learning approach for autonomous UAV navigation in indoor_drone_navigation environments. Our method combines path planning with DRL to help drones achieve a series of sub target tasks through task decomposition, ultimately completing global navigation tasks. This technique significantly improves navigation performance compared to standard DRL algorithms.

This experiment is conducted within the XuanCe framework and utilizes the gym-pybullet-drones UAV simulation engine for drone simulation.

• XuanCe Framework - A reinforcement learning framework used for algorithm implementation.
• gym-pybullet-drones - A drone simulation environment built on PyBullet.

Open terminal and type the following commands, then a new conda environment for xuance with drones could be built:

conda create -n xuance_drones python=3.10
conda activate xuance_drones
pip install xuance  

git clone https://github.com/utiasDSL/gym-pybullet-drones.git
cd gym-pybullet-drones/
pip install --upgrade pip
pip install -e .  # if needed, `sudo apt install build-essential` to install `gcc` and build `pybullet`

After the installation ends you can activate your environment with:

source activate xuance_drones

To initiate training using the DDPG algorithm, execute the following command:

python ddpg.py --device "cuda:0" --test 0 --seed 0 

To initiate training using the SAC algorithm, execute the following command:

python sac.py --device "cuda:0" --test 0 --seed 0 

When the training is complete, the data and lines can be observed in tensorboard.

tensorboard --logdir logs

python ddpg.py --seed 0 --test 1 --test_episode 1 --model_folder your_model_folder

or

python sac.py --seed 0 --test 1 --test_episode 1 --model_folder your_model_folder