search

osrf / subt

This repostory contains software for the virtual track of the DARPA SubT Challenge. Within this repository you will find Gazebo simulation assets, ROS interfaces, support scripts and plugins, and documentation needed to compete in the SubT Virtual Challenge.
Other
305 stars 98 forks source link

Map decimation error? #983

Open peci1 opened 3 years ago

peci1 commented 3 years ago

I had a local simulation running which was sending data to the /cloud topic. I saw this in the bridge container:

[pcl::VoxelGrid::applyFilter] Leaf size is too small for the input dataset. Integer indices would overflow.

I assume it is a warning from the map decimator.

I have no more details, though - mostly, I don't have any message from the /cloud topic as I came to an already stopped simulation.

mjcarroll commented 3 years ago

That does look like it would be from the map decimation. I believe that the leaf size is currently set to 0.1 meters: https://github.com/osrf/subt/blob/ce5905f9c611ae19f17b0e85975b19ca5cbe654b/subt_ros/src/cloud_downsample.cpp#L99

So it would take a large cloud in physical dimensions (not number of points) to trigger this warning. I think that would mean that you have a point outside INT_MAX 0.1 or INT_MIN 0.1. Is there any chance there you had some erroneous data in there?

peci1 commented 3 years ago

I don't say it's impossible, but the map is generally good and I've never noticed this wrong data. However, if it's a point faaaar away, it's a question whether it would ever been seen in rviz...

mjcarroll commented 3 years ago

However, if it's a point faaaar away, it's a question whether it would ever been seen in rviz...

Maybe we could just do a pre-filter for the INT_MAX 0.1 and INT_MIN 0.1 to guarantee that there aren't any points at millions of meters away.

peci1 commented 3 years ago

sounds good

peci1 commented 3 years ago

It happened again. Here are properties of the published cloud.

Num points: 642918 X range: -53 .. 381 Y range: -202 .. 27 Z range: -27 .. 29

peci1 commented 3 years ago

The cloud also has 19 channels (x,y,z included; point step 124). Total data length is 79721832 bytes.