search

YJZLuckyBoy / liorf

This repo is modified based on LIO_SAM, which remove the feature extraction module and makes it easier to adapt your sensor.
MIT License
388 stars 75 forks source link

Advice on how to improve performance #23

Open JesusSilvaUtrera opened 1 year ago

JesusSilvaUtrera commented 1 year ago

First of all, thanks for the great work!

I am trying to use this repo to perform 3D SLAM with my robot, but I am facing a problem that sometimes it gets lost and finishes creating a bad map. I am using it indoors, so I am not using any GPS data, even though the IMU provides it.

I have looked into my IMU data to see what happens when it gets lost, and I haven't discovered any anomaly in the magnetometer (I first thought it was because the robot was passing near high voltage cables for example) or in the orientation values. The thing is that usually the places where it gets lost are narrow places, like a corridor or when changing between rooms, so I don't know if this is a bug in liorf because it is optimized to be used outdoors or something.

Also, if you could give me some advice on which parameters should I tune to improve performance, I would really appreciate it.

Thanks in advance for the help!

JesusSilvaUtrera commented 1 year ago

For your information, I am using Robosense RS-16 LiDAR

JesusSilvaUtrera commented 1 year ago

@YJZLuckyBoy I also have another question. I am trying to use liorf with Navigation Stack, transforming the 3D map to an Occupancy Grid Map but using the 3D localization that liorf provides. The problem I am facing is that the '2D Pose Estimate' plugin is not working with liorf, and I don't know how to fix this. Do you have any advice on how could I do this? I have looked into liorf_localization repo and I have seen that it only supports the 3D maps that liorf creates, so I don't know how to handle this with 2D maps.