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This repository has been archived by the owner on Nov 8, 2023.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

Latest commit

 

History

History

orca_bringup

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Launch files

Bring up all core ROV and AUV nodes, including SLAM and Nav2. This is the primary launch file, but don't launch this file directly: instead launch sim_launch.py or topside_launch.py.

Launch ROV or AUV simulation in Gazebo (SITL). Includes bringup.py.

Launch device nodes on the sub (HITL).

Launch ROV or AUV topside nodes (HITL). Includes bringup.py.

Launch a SLAM simulation in Gazebo (SITL). Does not include bringup.py. Useful for testing the SLAM software.

Launch just the nodes required for stereo camera calibration (HITL).

Launch device and test nodes on the sub. Useful for simple hardware tests.

Video pipelines

There are 2 distinct stereo pipelines right now. Ideally the SITL and HITL pipelines would share more code. This is an area of active development.

SITL

The simulations use Gazebo camera sensors to generate uncompressed 800x600 images in an ideal stereo configuration. The Gazebo camera sensor provides for distortion (used in slam_test_launch) but not rectification. The cameras generate images at 30fps but the SLAM software is able to process the images at 15-20fps.

HITL

The sub generates three H264 streams: one 1920x1080 30fps stream for the forward camera and two 1640x1232 20fps streams for the left and right down-facing cameras. The 4X higher resolution for the left and right cameras as well as the required decoding, undistortion and rectification slows the SLAM pipeline down to ~2.5fps. The SLAM software is currently slowed down to 1fps to reduce CPU usage during testing.

orb_slam2_ros

The following changes were made on the clyde_h264_stereo branch of orb_slam2_ros:

  • Rotate the point cloud to support down-facing cameras
  • Decode, undistort and rectify H264 streams in proc
  • Some bug fixes (don't undistort twice, use camera_info.p instead of camera_info.k for intrinsics)