Re: Realtime Map Merger Improvements

Posted by derektan1995 on Sep 10, 2021; 5:15am
URL: http://official-rtab-map-forum.206.s1.nabble.com/Realtime-Map-Merger-Improvements-tp8200p8565.html

Hi Mathieu,

Thanks for your response, as always!

matlabbe wrote

It is more efficient to subscribe to mapData from rtabmapviz than the raw point clouds. The mapData contains only data (compressed) from the latest frame added to map. rtabmapviz will accumulate them (in its cache) and reconstruct the 3d map on its side. On loop closure, the mapData contains the modified graph (which is only some KB of data) and rtabmapviz re-correct the position of each cloud previously received in its cache without having to re-download data of each node.

Interesting, I did not know that the /mapData topic only contains the lastest frame added to the map. Let's imagine a scenario where, say, the computer hosting the multi-robot RViz is unable to subscribe to a robot's /mapData topic for an extended period of time. After that duration, the computer (i.e. more specifically, the rtabmap_ros/MapCloud plugin) manages to re-subscribe to the robot's /mapData topic. Since we have not been subscribing to the earlier frames, and that we are now re-subscribing to the latest frame, will that result in a 'big hole' in the map displayed on RViz? Are there ways for the MapCloud plugin to recover the missing nodes during the period of 'loss-comms', given that the robot still have all the nodes present in its rtabmap memory?

Thanks as always,
Derek