Deep reinforcement learning for drone precision landing, docker container for simulation in Gazebo-ROS2 dashing with PX4-Autopilot controller.
To start the docker run:
- For ROS2 Dashing:
./run_docker.sh run
Open 2 terminals and run the docker in each one of them, as explained above.
In the first one run:
ros2 run px4_ros_extended gazebo_runner.py --test --headless
In the second one run
cd /src/shared
./launch_test_ddpg.sh 0 small 6 3 4500
The simulation video for this implementation can be found at this link in which the landing platform is moving linearly. Another simulation video link shows landing of UAV on still platform.
@article{
author = {Alejandro Rodriguez-Ramos and Carlos Sampedro and Hriday Bavle and Paloma de la Puente and Pascual Campoy},
doi = {10.1007/s10846-018-0891-8},
issn = {15730409},
issue = {1-2},
journal = {Journal of Intelligent and Robotic Systems: Theory and Applications},
title = {A Deep Reinforcement Learning Strategy for UAV Autonomous Landing on a Moving Platform},
volume = {93},
year = {2019},
}
The initial code for the DDPG algorithm and Framework has been taken from this and this github repository.