Acknowledgements

This repository is a simplified version of the REMARO Summer School Delft 2022 - Underwater robotics hackathon. For more info, please visit: remaro-network/tudelft_hackathon

Setup

Tested with:

• Ubuntu 22.04
• ROS2 Humble
• Gazebo (Ignition) Garden
• ArduPilot (Sub-4.1)
• ardupilot_gazebo plugin
• mavros2
• remaro_world
• bluerov2_ignition

Installation

Disclaimer Running docker images with graphical user interface is a little bit trick and might not work in all systems. We tested in a system with ubuntu 22.04 and with a NVIDIA gpu. It might not work on systems with AMD gpus, and on MAC.

Install prerequisites to run with docker

• Install docker on your machine. You can find instructions here
• Allow non-root users to manage docker. Instructions here
• Install VSCode. Instructions here
• Install nvidia-docker(only needed if you have a nvidia GPU)

Run it with docker via CLI

Create docker network:

sudo docker network create ros_net

Run Ignition simulation + ardupilot SITL:

If you a NVIDIA GPU:

xhost +local:root
sudo docker run -it --rm --name ignition --net ros_net -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix:ro --gpus all ghcr.io/remaro-network/tudelft_hackathon:nvidia ros2 launch tudelft_hackathon bluerov_ign_sim.launch.py ardusub:=true mavros_url:='bluerov:14551'

If you have an AMD GPU:

xhost +local:root ;
sudo docker run -it --rm --name ignition --net ros_net -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix:ro --device=/dev/dri --group-add video  ghcr.io/remaro-network/tudelft_hackathon:non-nvidia ros2 launch tudelft_hackathon bluerov_ign_sim.launch.py ardusub:=true mavros_url:='bluerov:14551'

If you have an Intel GPU:

xhost +local:root ;
sudo docker run -it --rm --name ignition --net ros_net -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix:ro --device=/dev/dri:/dev/dri  ghcr.io/remaro-network/tudelft_hackathon:non-nvidia ros2 launch tudelft_hackathon bluerov_ign_sim.launch.py ardusub:=true mavros_url:='bluerov:14551'

Development with docker via cli

To add your modifications into the docker images you need to rebuild the relevant docker images. In this case, run the build-dev-images.sh script to rebuild them. And make sure to substitute in the docker run commands the images from rezenders to you local images. I.e: rezenders/ignition:hackathon -> ignition:hackathon-dev and rezenders/ros-foxy-hackathon -> ros-foxy-hackathon:dev

Run it with docker with VSCode

Check instructions here

Run it locally

Simulation

Before running anything you need to source the workspace. With this command:

source ~/tudelft_hackathon_ws/install/setup.bash

Or you can add that to the ~/.bashrc file to prevent needing to source everytime.

echo "source ~/tudelft_hackathon_ws/install/setup.bash" >> ~/.bashrc

Don't forget to re-open your terminal after altering the ~/.bashrc file.

In one terminal run ardusub SITL:

  sim_vehicle.py -L RATBeach -v ArduSub  --model=JSON --console

In another terminal run the simulation + mavros + agent:

 ros2 launch tudelft_hackathon bluerov_bringup.launch.py simulation:=true ardusub:=false

Explanation

Simplified system architecture:

The system was designed to be used both with a real or simulated BlueROV2. When used with a simulation, the left nodes are deployed. And when used with the real robot the right nodes are deployed.The agent and MAVROS nodes are always deployed.

First, let's take a look on how the real BlueROV2 works, then we see how the simulation is setup to mimic it.

Simulated BlueROV2

Simulated BlueROV2: To simulate the BlueROV2 we are using Gazebo (Ignition). Unfortunately, until the moment of writing this readme, there is no sonar plugin for Ignition. Thus, we are using a lidar plugin instead, configured to have the same speed and measurement range as the Ping360 sonar. The bluerov2 model is here and the bluerov2 model with a lidar is here. The world being used for the simulation can be found here, note that there is a buoyancy plugin that sets the "water" density to 1000kg/m3.

Simulated "sonar" bridge: In order to have access to the simulated lidar data with ROS2 we run a bridge between ignition transport and ROS2. Something like this:

Via command line:

ros2 run ros_gz_bridge parameter_bridge lidar@sensor_msgs/msg/LaserScan@ignition.msgs.LaserScan -r /lidar:=/scan

With launch file:

package='ros_gz_bridge',
    executable='parameter_bridge',
    arguments=['lidar@sensor_msgs/msg/LaserScan@ignition.msgs.LaserScan'],
    remappings=[('/lidar','/scan')],
    output='screen'

Note that the topic where the lidar data is published has the same name (/scan) and type (sensor_msgs/LaserScan) as the topic published by the ping360 driver.

ArduSub Sofware In The Loop (SITL): Since when running the simulation we don't have a board with an autopilot installed, we simulate Ardusub as a SITL.

Ardupilot gazebo plugin: Bridge between Ignition and Ardusub. More info can be found here.

MAVROS: The only difference is that for the simulation we need to use a different fcu_url. In this case, udp://:14551@:14555.

Agent: Since all the interfaces are the same, the agent nodes are the same for both simulation and the real robot.