Skip to content
This repository has been archived by the owner on Nov 8, 2023. It is now read-only.

Latest commit

 

History

History
60 lines (47 loc) · 2.81 KB

README.md

File metadata and controls

60 lines (47 loc) · 2.81 KB

Orca3 Localization

Perhaps the biggest challenge in underwater robotics is localization:

  • radio signals (e.g., GPS) are absorbed rapidly in water, particularly seawater
  • you often can't see the seafloor
  • sonar is useful, but it is typically big, expensive and power-hungry (although this is improving)
  • inexpensive IMUs aren't good for more than a few meters, and the good ones are big and expensive

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

  • fiducial_vlam looks for ArUco markers in the environment using the forward-facing camera. This is useful for test environments like pools.
  • orb_slam2_ros looks for ORB features along the seafloor. It is configured to use a custom down-facing stereo camera. This works well in simulation, but it has not been tested in the field.

FiducialLocalizer

The FiducialLocalizer node does the following:

  • subscribe to /fiducial_observations (a list of marker observations) and /forward_camera/camera_pose (the pose of the camera generated by solvePnP or GTSAM)
  • reject poses generated by markers that are too far away (~2m). This avoids unstable solutions.
  • calculate the map->base_link transform
  • publish the map->odom transform

No transforms are published until a marker is observed. After the first observation transforms are published at the target rate with updated timestamps, even if they are old.

Parameters

Parameter Type Default Notes
map_frame_id string map Map frame id
odom_frame_id string odom Odom frame id
camera_frame_id string camera_frame Camera frame id
publish_rate int 20 How often to publish transforms, Hz
wait_for_transform_ms int 500 How long to wait for the odom->base transform, ms
transform_expiration_ms int 1000 How far ahead to post-date the map->odom transform, ms
good_pose_distance double 2.0 Maximum distance for a good marker observation, m

OrbSlam2Localizer

The OrbSlam2Localizer node does the following:

  • subscribe to /camera_pose
  • calculate the map->base_link transform
  • publish the map->odom transform

The systems starts as soon as there's a good pose. The simulated world has a depth of 4m and the mapping starts immediately.

Parameters

Parameter Type Default Notes
map_frame_id string map Map frame id
odom_frame_id string odom Odom frame id
camera_frame_id string camera_frame Camera frame id
publish_rate int 20 How often to publish transforms, Hz
transform_expiration_ms int 1000 How far ahead to post-date the map->odom transform, ms