search

luxonis / rae-ros

Implementation of RAE ROS and gazebo stack
MIT License
32 stars 13 forks source link

SLAM clients #37

Open sskorol opened 1 year ago

sskorol commented 1 year ago

Hi,

Do you have any instructions on how to use rtabmap on the client side? I enabled it via ros2 launch rae_bringup bringup.launch.py enable_slam_toolbox:=false enable_rtabmap:=true.

Screenshot from 2023-10-17 15-16-03

I can add a mapcloud in Rviz. But I don't understand what I see (as it doesn't reflect any real shape in front or back of the robot) and how it could help me to do SLAM. I mean, the map seems static even if I move the robot. IMU doesn't reflect any orientation changes. Do you have rviz config for slam-toolbox and rtabmap? I also wonder how you load the robot's description on PC when meshes are physically on RAE? Please advise if you're using some other toolset for SLAM testing.

Serafadam commented 1 year ago

Hi, in next upcoming updates we're going to add improvements to Odometry publishing, as well as changes to bringup of some Nodes which should improve performance of SLAM algorithms. Additionally, in upcoming weeks there should be major changes with the robot's firmware leading to large performance improvements. When it comes to publishing pointclouds, depending on network configuration this can be taxing on the ROS2 side, one of the ways to mitigate that is to tune DDS parameters as in official ROS2 Guides. One other change that could improve publishing is to use foxglove_bridge interface for data visualization. As for the robot meshes, you could download and build the ROS packages on your local computer, after sourcing Rviz should be able to detect them.

sskorol commented 1 year ago

That's great! Let me know when it's updated, please.