Google Cartographer 3D map for localization

Hello, I want to use google cartographer 3d pure localization in my project. I have Velodyne VLP-16 LIDAR. First I will create a 3D Map for localization. So cartographer will use this pbstream file for localization.

In 3D, cartographer does not publish a 2D occupancy grid map so i will create 2D map with slam-gmapping for ROS navigation Stack. I plan to use cartographer only for localization instead of AMCL.

Localization and Navigation will be made on different maps. It is a problem.

My question is ; how I can make Cartograpger 3D localization and ROS Navigation Stack work together?

Asked by zeynep on 2019-06-12 07:18:58 UTC

Comments

Answers

You should use pbstream_map_publisher and/or occupancy_grid_node to create the occupancy grid. If you are running pure localization then your maps will be the same since they will both be generated from the same pbstream.

Asked by JustinBlack02 on 2019-06-15 13:17:28 UTC

Comments

Thanks for your reply but accroding to https://github.com/googlecartographer/cartographer_ros/issues/330 it is not a 2D Occupancy Grid Map that ROS Navigation Stack can use, it is a X-ray map . According to damonkohler answer, they include all range data projected into a 2D plane. X-rays are a 2D representation of vertical density that's useful for humans, but probably not for robots. So it is useless for Navigation Stack.

Asked by zeynep on 2019-06-15 14:17:21 UTC

Your entire problem is self inflicted. You localize in 3D and navigate in 2D. Why don't you make your dimensions match? If you navigate in 2D then localize in 2D and use the occupancy grids. If you really want to localize in 3D then you should query the submaps for their point clouds, convert those to e.g. octomaps for 3d navigation or slice along the z axis in the domain your robot will navigate and project the resulting cloud down to 2d to convert that to an occupancy grid.

Asked by JustinBlack02 on 2019-06-15 15:16:08 UTC

I will try to convert 3D pbstream file to point cloud data. I will create an elavation map in 2D from this point cloud data.

Asked by zeynep on 2019-06-17 03:49:45 UTC

Please mark this answer as accepted or add your own answer for future users.

Asked by JustinBlack02 on 2019-06-17 08:18:54 UTC