Orca3

Orca3 is a set of ROS2 packages that provide basic AUV (Autonomous Underwater Vehicle) functionality for the BlueRobotics BlueROV2.

Orca3 includes a base_controller node that responds to 3D 4DoF /cmd_vel messages and generates thrust and odometry. The BlueROV2 barometer sensor is used to hover at the target depth. See orca_base for details.

Orca3 supports two SLAM (Simultaneous Localization And Mapping) systems:

fiducial_vlam looks for ArUco markers in the environment using the forward-facing camera.
orb_slam2_ros looks for ORB features along the seafloor. It is currently configured to use a custom down-facing stereo camera, but the underlying ORB-SLAM2 system also works with depth and monocular cameras.
Localization details

Orca3 uses the ROS2 Navigation2 framework for mission planning and navigation. Several Nav2 plugins are provided to work in a maritime environment:

straight_line_planner_3d
pure_pursuit_3d
progress_checker_3d
goal_checker_3d
Nav2 plugin details

Orca3 is simulated in Gazebo 11. Several Gazebo plugins are provided to support a maritime environment:

OrcaDragPlugin
OrcaBuoyancyPlugin
OrcaThrusterPlugin
OrcaBarometerPlugin
Gazebo plugin details

Installation

Orca3 has been tested on Ubuntu 20.04 and ROS2 Foxy. See the Dockerfile for installation instructions.

Simulation with fiducial_vlam

In a terminal run:

cd ~/ros2/orca_ws
source /opt/ros/foxy/setup.bash
source install/local_setup.bash
export GAZEBO_MODEL_PATH=${PWD}/install/sim_fiducial/share/sim_fiducial/models
source /usr/share/gazebo/setup.sh
ros2 launch orca_bringup sim_launch.py gzclient:=True rviz:=True slam:=vlam world:=ping_pong

This will start Gazebo, Rviz2 and a bunch of ROS2 nodes. If everything started correctly, you will see Gazebo and Rviz2 GUIs like this:

The surface of the water is at Z=0 and the AUV will be sitting at the surface. There's one ArUco marker directly in front of the AUV to provide a good initial pose. There are two more AUV markers 2 meters below the surface, facing each other 10 meters apart.

Execute a mission by calling the FollowWaypoints action in a second terminal:

# Aligns with markers on the ping-pong map
ros2 action send_goal /FollowWaypoints nav2_msgs/action/FollowWaypoints "{poses: [\
{header: {frame_id: 'map'}, pose: {position: {x: 4, z: -2}}},\
{header: {frame_id: 'map'}, pose: {position: {x: -4, z: -2}}},\
{header: {frame_id: 'map'}, pose: {position: {x: 4, z: -2}}},\
{header: {frame_id: 'map'}, pose: {position: {x: -4, z: -2}}},\
]}"

This mission will dive to -2 meters, then ping-pong between the two submerged markers. The AUV will only trust a marker pose if the marker is within 2 meters, so it will run dead-reckoning ~7 meters between the markers. As the AUV approaches a marker and gets a good pose, the dead-reckoning error is added to the map->odom transform.

Simulation with orb_slam2_ros

In a terminal run:

cd ~/ros2/orca_ws
source /opt/ros/foxy/setup.bash
source install/local_setup.bash
export GAZEBO_MODEL_PATH=${PWD}/install/sim_fiducial/share/sim_fiducial/models
source /usr/share/gazebo/setup.sh
ros2 launch orca_bringup sim_launch.py gzclient:=True rviz:=True slam:=orb world:=empty

This Rviz2 configuration is tuned for orb_slam2_ros:

The surface of the water is at Z=0 and the AUV will be sitting at the surface. The world contains a rocky seafloor 4 meters below the surface, indicated by the blue marker in Rviz2. The pointcloud shows the ORB features recognized by orb_slam2_ros.

Execute a mission by calling the FollowWaypoints action in a second terminal:

# Run a 6m x 6m box 1.5m above the seafloor, will do a loop closure
ros2 action send_goal /FollowWaypoints nav2_msgs/action/FollowWaypoints "{poses: [\
{header: {frame_id: 'map'}, pose: {position: {x: 3, y: 3, z: -2.5}}},\
{header: {frame_id: 'map'}, pose: {position: {x: -3, y: 3, z: -2.5}}},\
{header: {frame_id: 'map'}, pose: {position: {x: -3, y: -3, z: -2.5}}},\
{header: {frame_id: 'map'}, pose: {position: {x: 3, y: -3, z: -2.5}}},\
{header: {frame_id: 'map'}, pose: {position: {x: 3, y: 3, z: -2.5}}},\
]}"

This mission will dive to -2.5 meters and move in a 6m x 6m square. The AUV has a good view of the seafloor, and orb_slam2_ros runs at the camera frame rate so the map->odom transform is published at 30fps. There is some odometry error which accumulates in the map->odom transform. You should notice a loop closure when the square is completed. The adjustment is very small.

[orb_slam2_ros_stereo-12] Loop detected!
[orb_slam2_ros_stereo-12] Local Mapping STOP
[orb_slam2_ros_stereo-12] Local Mapping RELEASEStarting Global Bundle Adjustment
[orb_slam2_ros_stereo-12] 
[orb_slam2_ros_stereo-12] Global Bundle Adjustment finished
[orb_slam2_ros_stereo-12] Updating map ...
[orb_slam2_ros_stereo-12] Local Mapping STOP
[orb_slam2_ros_stereo-12] Local Mapping RELEASE
[orb_slam2_ros_stereo-12] Map updated!

Packages

Future work

This project is under heavy development as of January 2021. Planned work:

Field tests!
Nav2 recovery plugin(s) for underwater environments

keevindoherty / orca3