maayanshvo/parrot-ar-drone-env

A Parrot AR Drone Environment for RL(tested on Python3)

A ROS package to setup a Parrot AR drone OpenAI Gym environment for Reinforcement Learning experiments

Building Workspace in python3:

After cloning the repository, follow the following steps:

cd parrot-ar-drone-env
wstool up
rosdep install --from-paths src --ignore-src -y -r
catkin_make --cmake-args \
            -DCMAKE_BUILD_TYPE=Release \
            -DPYTHON_EXECUTABLE=/usr/bin/python3 \
            -DPYTHON_INCLUDE_DIR=/usr/include/python3.6m \
            -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.6m.so

A sample test file can be found at src/parrot_ardrone_rl/scripts. Currently I am subscribing to gt_pose, gt_vel and front_camera/image_raw topics as the observation tuple (Dimensions are [(13,),(360,640,3)]).

For the action space, I am publishing cmd_vel with angular x and y velocities set to zero. You can edit these settings in parrot_gym/parrotdrone_env.py.

The tasks are defined in parrot_gym/parrotdrone_tasks.

Note: I recommend creating a virtual python3 environment and using the test file as a guide to build your own training/testing project.