ROS2 various QOL updates (preparing new binary release)

[matlabbe]

TODO:

• doublecheck if there is no regression for 3D LiDAR + deskewing. Add vlp16 example with IMU and/or T265 (if working on ros2).
  • https://github.com/introlab/rtabmap_ros/blob/master/rtabmap_examples/launch/test_velodyne_t265_deskewing.launch
  • https://github.com/introlab/rtabmap_ros/blob/master/rtabmap_examples/launch/test_velodyne_d435i_deskewing.launch
  • https://github.com/introlab/rtabmap_ros/blob/master/rtabmap_examples/launch/test_velodyne.launch
• Convert demo_husky.launch
• Convert demo_appearance_mapping.launch. Won't do: there is no point to this online. Camera node required for this has been pushed to rtabmap_legacy in ros1. I think if we create a new ROS node wrapper for Camera class of rtabmap, we may revive this example.
• Convert demo_find_object.launch
• Convert demo_multi-session_mapping.launch
• Convert demo_isaac_carter_navigation.launch: Won't do The lidar format cannot be easily used out-of-the-box to appropriately do deskewing with the current deskewing approach. The "assembled" 360 scans don't have timestamp offset field set in the point cloud for every point by default. More work to make a proper demo with lidar would need to be done. For a VSLAM demo using Isaac simulator, we may check to add it in the future (maybe at the same time than an integration demo with Isaac vslam nodes). Update: I added a VSLAM demo working with basic Isaac sim navigation examples.
  • Note that we have to enable manually the right camera (and decrease resolution to save computation, same for left camera):
    
  • For the iw_hub_warehouse example, the right camera doesn't include Tx value, so depth image cannot be generated correctly. To fix that, in the Action Graph of the front hawk camera, disable left and right camera info publishers, then add a "ROS2 Camera Info Helper" with those configs:
    
  • Another issue, it seems after reducing Physic->Steps per second to 30 Hz instead of 60 Hz, the cameras are not synchronized anymore (even if they are published with exact same stamp):
    
• Convert demo_robot_mapping.launch
• Convert demo_stereo_outdoor.launch
• Convert demo_unitree_quadruped_robot.launch : Won't do: the unitree gazebo example has not been ported to ros2. See also this repo (https://github.com/IntelligentRoboticsLabs/go2_robot) for go2 robot with nav2, though not available in sim yet.
• quadrupedal robots VSLAM demo: https://github.com/chvmp/champ: current issue: tf2::Buffer::lookupTransform seems blocking Listener's thread when subscribing to an image topic on high CPU load ros2/geometry2#727

Converting rosbag from ROS1 to ROS2: https://gitlab.com/ternaris/rosbags
Fixing leading slash in topic frames: https://github.com/cartographer-project/cartographer_ros/blob/master/cartographer_ros/scripts/remove_leading_slashes.py

Converted rosbags https://drive.google.com/drive/folders/1FMz-3xTI4-ouKxOUXlRvyty0jmaUL7iQ?usp=drive_link

New Demos

Outdoor Stereo VSLAM

ros2 launch rtabmap_demos stereo_outdoor_demo.launch.py

Indoor 2D LiDAR and RGB-D SLAM

ros2 launch rtabmap_demos robot_mapping_demo.launch.py rviz:=true rtabmap_viz:=true

Multi-Session Indoor 2D LiDAR and RGB-D SLAM

ros2 launch rtabmap_demos multisession_mapping_demo.launch.py

Find-Object with SLAM

ros2 launch rtabmap_demos find_object_demo.launch.py

Turtlebot4 Nav2, 2D LiDAR and RGB-D SLAM

ros2 launch rtabmap_demos turtlebot4_sim_demo.launch.py

Turtlebot3 Nav2 and 2D LiDAR SLAM

ros2 launch rtabmap_demos turtlebot3_sim_scan_demo.launch.py

Turtlebot3 Nav2 and RGB-D SLAM

ros2 launch rtabmap_demos turtlebot3_sim_rgbd_demo.launch.py

Turtlebot3 Nav2, 2D LiDAR and RGB-D SLAM

ros2 launch rtabmap_demos turtlebot3_sim_rgbd_scan_demo.launch.py

Champ Quadruped Nav2, Elevation Map and VSLAM

ros2 launch rtabmap_demos champ_sim_vslam.launch.py

Clearpath Husky Nav2, 2D LiDAR and RGB-D SLAM

ros2 launch rtabmap_demos husky_sim_scan2d_demo.launch.py

Clearpath Husky Nav2, 3D LiDAR and RGB-D SLAM

ros2 launch rtabmap_demos husky_sim_scan3d_demo.launch.py

Clearpath Husky Nav2, 3D LiDAR Assembling and RGB-D SLAM

ros2 launch rtabmap_demos husky_sim_scan3d_assemble_demo.launch.py

Isaac Sim Nav2 and Stereo SLAM

ros2 launch rtabmap_demos isaac_sim_vslam_demo.launch.py

Isaac Sim Nav2 and RGB-D VSLAM

ros2 launch rtabmap_demos isaac_sim_vslam_demo.launch.py stereo:=false vo:=rtabmap

[TannerGilbert]

Hey @matlabbe would it be possible to add PointCloud assembling to the rtabmap.launch.py similar to how it's available in rtabmap.launch for ROS 1?

[matlabbe]

@TannerGilbert It depends on what you want to do. I created new examples with point_cloud_assembler, so they may be more appropriate for your use case. Even in ros1, I didn't use point cloud assembling in rtabmap.launch for a while, I refer people to VLP16 examples instead, which come with already tuned parameters and different approaches of deskewing. That is something that this PR will provide for ROS2.

[TannerGilbert]

@TannerGilbert It depends on what you want to do. I created new examples with point_cloud_assembler, so they may be more appropriate for your use case. Even in ros1, I didn't use point cloud assembling in rtabmap.launch for a while, I refer people to VLP16 examples instead, which come with already tuned parameters and different approaches of deskewing. That is something that this PR will provide for ROS2.

Ok makes sense. I'll switch to a format similar than the VLP16 demo for my launchfile then.

[matlabbe]

@TannerGilbert The new point cloud assembling example can be found here: https://github.com/introlab/rtabmap_ros/blob/ros2/rtabmap_examples/launch/lidar3d_assemble.launch.py

cheers,
Mathieu