openuav-ros-px4-tensorflow

This repo keeps the setup for a complete docker container using ansible and docker This dockerfile makes a image with the following installs
-openuav repo
-ros Kinetic
-px4/mavros/mavlink
-openai gym
-tensorflow

To build:

run: sudo nvidia-docker build --no-cache=true -t openuav-gz-tf -f Dockerfile.devel-gpu
Note: this will take a while to setup (~45min)

Once running you must do the following steps:

1. Build again. Note px4 will hang (a bug) so give it a couple seconds then end it
   cd ~/catkin_ws/ && catkin build
2. Rerun bashrc. When logging in bashrc is not automatically run so make sure to run it each time you log in
   source ~/.bashrc

To run with ansible:

add "localhost ansible_connection=local" to /etc/ansible/hosts file
Run the playbook. The ports are automatically changed to 75 85 so when you run gzserver use -p 85
ansible-playbook test-ansible.yml --extra-vars "outport =85 inport=75"

To add GUI application/Camera simulation support:

1. Log into the host machine physically (no ssh). Xhost will not allow changes remotely for security reasons
   

2. Run xhost +local: . Be aware of the security issues. This allows any user on the machine to have access to the xsession
   

3. Run the container with:
   docker run -dit \
   -v /tmp/.X11-unix:/tmp/.X11-unix <br /> -e DISPLAY=$DISPLAY <br /> openuav-gz-tf bash
   

4. This should make a container that forwards xapps to the machine (ssh -X will not push that to your remote machine. See below). This will also allow camera emulation inside of a ros simulation. For remote access to the camera, setup ssh server on conatiner and connect through a forwarded docker port.

Using local Xserver for camera support

On container, run the following

Xvfb :1 -screen 0 1600x1200x16  &
export DISPLAY=:1.0