Geometry Boost List: Re: [geometry] R-tree vs octomap

Subject: Re: [geometry] R-tree vs octomap
From: Adam Wulkiewicz (adam.wulkiewicz_at_[hidden])
Date: 2014-05-23 16:47:19

Hi Andrew,

Andrew Hundt wrote:
> I was hoping to find out a bit about the similarities and differences 
> in capability between R-tree and an octomap, which is a bit like a 
> volumetric or voxel base octree.
>
> Octomap:
> https://octomap.github.io/
>
> Does anyone have thoughts on the use cases and overlapping 
> advantages/disadvantages of an R-tree and octomap?

AFAIU it's an octree. The difference between an octree and rtree is that 
the first one partions space and stores the objects occupying some space 
in a node, and the second one gathers some close objects together and 
encloses them in nodes.

I think that from a user-point of view all space-indexes could allow 
doing the same things, not only those two.

Since octomap is used in robotics to store measurements from various 3D 
point-sample-based sensors it probably only allows to store Points. But 
since it's specialized, it may be somehow optimized to work well with 
some of the algorithms used in robotics/SLAM like the Iterative Closest 
Point algorithm which AFAIK is used in 3D SLAM.

Did you check it out? E.g. did you do some benchmarking?

Regards,
Adam