neobotix / neo_localization2

An AMCL-alternative for Indoor mobile robot navigation
MIT License
18 stars 15 forks source link

'Number of points too low' error when map size is close to ~1 km2 #17

Open jainRAV opened 2 years ago

jainRAV commented 2 years ago

Hi, thanks for this package.

This is related to this issue.

I was able to narrow down the symptoms of this error. I tried Gazebo (instead of Ignition) to narrow down the cause. When using AMCL, the CPU usage is a steady and no error occurs.

But when using neo_localization2, CPU sometimes spikes to 100% and the 'Number of points too low' error is shown. I am building the package with release arguments.

My setup: Two lasers in gazebo with relay to scan topic. Running this on latest Ignition+Gazebo docker file from OSRF. System is i7 9th gen with GTX1650ti. Also running nav2 parallely on same system.

Due to this, nav2 goal is also sometimes aborted.

Can you guess any reasons why this might be happening?

Thanks

jainRAV commented 2 years ago

I am using humble branch. It seems like aeed20fcf29d9f8a32520eecadebb6ba164c1768 is not synced. Can it be this?

padhupradheep commented 2 years ago

Thanks for the report!

Could you maybe post a link to your configuration and maybe the entire simulation setup?

After https://github.com/neobotix/neo_localization2/commit/aeed20fcf29d9f8a32520eecadebb6ba164c1768 it was synced to all distros except foxy. Also, we tested extensively on the simulation. We haven’t faced this issue.

jainRAV commented 2 years ago

Thanks. I will share the files in next few days.

automech-rb commented 2 years ago

Hi,

Got it working. I updated my map to a smaller maps and decreased "map_size" parameter. Sorry for disturbing you.

padhupradheep commented 2 years ago

Nice to know, what was the size of the map you intended to use before ?

automech-rb commented 2 years ago

It was a ~1km map so I was using the map_size about 20000.

padhupradheep commented 2 years ago

Interesting!

I'll leave this issue open, so that I can further investigate on it!