ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Navigation from PointCloud or Ocupancy Grid

asked 2020-06-03 18:52:22 -0500

czr gravatar image


I was able to apply rtabmap and build a occupancy grid and a point cloud for the ground plane and a pointcloud for the obstacles. Now a I want to use this data to navigate the robot autonomously. In this case, this would be outdoor navigation. I found the package move_base that seems to do that but I could not understand how to connect it to the data I already have.

So, I need some guidance on how to proceed next, which package implements navigation from stereo camera/3D ladar? Will I have to code this from scratch, if yes, which algorithms should I look into first?

edit retag flag offensive close merge delete


Did you see this tutorial?

matlabbe gravatar image matlabbe  ( 2020-06-04 10:07:55 -0500 )edit

Thank you. I followed it as it is. Then changed the openni_points topic for /rtabmap/cloud_obstacles, on the local_costmap_params.yaml file among other things but I always get the warning: The openni_points observation buffer has not been updated for x.xx seconds, and it should be updated every 0.50 seconds.

czr gravatar image czr  ( 2020-06-04 22:19:18 -0500 )edit

Is /rtabmap/cloud_obstacles published?

matlabbe gravatar image matlabbe  ( 2020-06-23 12:23:00 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted

answered 2020-06-03 22:06:50 -0500

Dragonslayer gravatar image

updated 2020-06-03 22:09:52 -0500

move_base is part of the ros navigation stack, which enables 2d navigation. You can use move_base and its global and local planners and costmaps. link You could start here, its a tutorial for the turtlebot, but all the files are on github and you can look them up. Should be mostly remapping topics and tuning the planners (specially the local planner, in the launchfiles and maybe some yaml file). Its made for indoor use though. I would serve the global planner a projection gridmap (rtabmap publishes this, I dont have the exact name handy), this is due to the fact that the navigation stack is 2d navigation. To take any kind of obstacle or robot height into consideration you have to "compress"/project the 3d data into the 2d gridmap, but as I said rtabmap delivers this cabability out of the box, rtabmap can also provide localization to correct odometry, just has to be put in localization mode (done in the launchfile). Also I only have seen rtabmap and navigation stack work with a prebuild database. SLAM with navigation stack and some sort of exploration algorithm/package I would only try in a second stage after the navigation with a prebuild database works, and this might involve some coding.

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2020-06-03 18:52:22 -0500

Seen: 1,279 times

Last updated: Jun 03 '20