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

storing 3d range data for localization

asked 2011-08-17 12:38:52 -0500

joschu gravatar image

As I understand it, the typical way to use the navigation stack with a laser range finder is this: drive around and collect range data, filling up the voxel grid, then save the 2D occupancy grid. Then the next day, you load this map, and amcl localizes the robot by comparing the new range data to the saved map.

It seems like localization would work better if you saved the 3D data and used that. Why doesn't the navigation stack use a 3d map?

edit retag flag offensive close merge delete


The range data you collect fills up an occupancy grid, not a voxel grid. But otherwise, yes.
Mac gravatar image Mac  ( 2011-08-17 16:58:25 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted

answered 2011-08-17 15:18:18 -0500

Zack gravatar image

updated 2011-08-17 15:22:27 -0500

It depends on your needs. 3D would provide would allow you to localize in a 3D space. However there are other considerations.

Firstly, localizing in the a 3D space would increase the complexity of the localization problem. It would thus impose a significant computation load on the system.

Secondly, it would also takes more memory just to store a 3D map as opposed to a 2D map. say you have 10 points for 1D, that would be 100 points for 2D and 1000 points for 3D.

Thirdly, the laser scanner only provides 2D data. So you might need to get a tilting mechanism to be able to build a 3d pointcloud. Or use other kind of sensors such as a kinect camera.

Most of the time when you are driving you just need to know your 2D location. It's nice to know what's your height with reference to AMSL; and you can install a fancy altimeter in your car. However you probably won't be looking at it much, simply because you don't really need the data.

So at the end of the day, it really depends on your needs.

edit flag offensive delete link more


I would add that the main idea behind the current system for localization has been developed more than 10 years ago when having a laser was already a big step over sonars. I can imagine in the future people will be dealing with 3D localization when, for example, an outdoor robot goes over hills.
Lorenzo Riano gravatar image Lorenzo Riano  ( 2011-08-17 21:41:41 -0500 )edit

Question Tools


Asked: 2011-08-17 12:38:52 -0500

Seen: 412 times

Last updated: Aug 17 '11