Comments (12)
You are right. Rviz definitely will use too much RAM if working too long. There are two method may help:
-
set the Decay time parameter of the topic cloud_registered in Rviz to a smaller one like 100 seconds.
-
The publish_dense only control the rviz shows, it will not influence the pcd saving. so you can set it to 0 to save the rviz.
May I ask how long is the duration of your bag usually? I am a little afraid that the pcd saving function may meet problem if the duration is too long.
from fast_lio.
Hey @XW-HKU , thanks for the quick answer. I tried changing the decay time, it does help somewhat yes so I will use it. My bag duration in most cases is around 15-20 min long. Yesterday while testing some more I noticed that sometimes the publish rate of the cloud_registered topic varies and is sometimes less than 10hz ( which is the lidar frequency ). Is that due to there being a lot of points to process in a very little time so the publishing is slowing down due to all the calculation or is there some other reason for this?
from fast_lio.
Hey @XW-HKU , thanks for the quick answer. I tried changing the decay time, it does help somewhat yes so I will use it. My bag duration in most cases is around 15-20 min long. Yesterday while testing some more I noticed that sometimes the publish rate of the cloud_registered topic varies and is sometimes less than 10hz ( which is the lidar frequency ). Is that due to there being a lot of points to process in a very little time so the publishing is slowing down due to all the calculation or is there some other reason for this?
The runtime performance of fastlio will not be influenced by the duration, but rviz will eat much computation resources and also the pcd saving function. You can close the rviz and disable the pcd saving function and test again to varify if the runtime performance is influenced.
from fast_lio.
And what's your lidar type, computer and its cpu?
from fast_lio.
So, after turning off both RViz and the pcd saving function, RAM usage no longer goes up and stays pretty much stable, I also tested with just pcd saving but not RViz and saw that it still takes RAM but a lot less than if RViz is working as well. The sensor I am using is an OS1-128 from Ouster and the pc has an 8-core ARM v8.2 64-bit CPU with 32GB RAM
from fast_lio.
But I noticed also that when switching between the publish_dense
option for the same exact bag in ros, one scan ended up with 2.1mil points without the dense scan published, and with over 20 mil points when i set the publish_dense
back to 1, which was a bit weird to me cause I thought it shouldn't make a difference for the end map
from fast_lio.
But I noticed also that when switching between the
publish_dense
option for the same exact bag in ros, one scan ended up with 2.1mil points without the dense scan published, and with over 20 mil points when i set thepublish_dense
back to 1, which was a bit weird to me cause I thought it shouldn't make a difference for the end map
The dense_publish is the switch to control the point number of the rostopic \cloud_registered. I am a bit confused about the "end map" you mentioned. What's that? If you are talking about the pcd file containing every registered scans, the topic \cloud_registered shouldn't change the pcd point number (if pcd's point number changed, there must be a bug I need fix, please tell me). As for the localization, it only need a sparse map which is not published and irrelevant to publish_dense parameter.
from fast_lio.
@XW-HKU yes, the dense_publish for me changed the number of points in the end pcd file that is being saved ,in one case it was 729MB file with about 21mil points and with dense_publish off it was 66MB with around 2.1mil points for the exact same bag file
from fast_lio.
@XW-HKU yes, the dense_publish for me changed the number of points in the end pcd file that is being saved ,in one case it was 729MB file with about 21mil points and with dense_publish off it was 66MB with around 2.1mil points for the exact same bag file
Thank you very much for finding this problem, I will fix it ASAP.
from fast_lio.
@XW-HKU thank you, perhaps the problem is in publish_world_frame
function the size of the laserCloudWorld
is determined by whether you have dense_publish enabled or not
from fast_lio.
@XW-HKU yes, the dense_publish for me changed the number of points in the end pcd file that is being saved ,in one case it was 729MB file with about 21mil points and with dense_publish off it was 66MB with around 2.1mil points for the exact same bag file
I just test the conditions the parameter scans_dense_enable is ON/OFF, but the result in my computet shows that the two pcd file are totally same size. So could you please update to the newest code and check again?
from fast_lio.
Hi, @XW-HKU, thank you for fixing it, I pulled the latest code and tried it again and now the pcd is the same size with and without dense_pub_en
from fast_lio.
Related Issues (20)
- fatal error: fast_lio/Pose6D.h: No such file or directory. HOT 1
- fastlio+livox HOT 1
- velodyne not fit into this project. HOT 2
- About NCLT Dataset ground truth HOT 1
- Downsampling Error HOT 1
- Livox MID-360 FAST_LIO2 建图过程中地图消失问题
- Gmapping navigation HOT 3
- The fineness of point cloud display in rviz HOT 1
- 里程计输出频率问题? HOT 1
- FAST_LIO uses only one core of the CPU HOT 1
- Is IMU optional? HOT 3
- 话题/Laser_map为空 HOT 3
- Some Clarification in Code HOT 2
- Is livox_ros_driver2 required at all? HOT 3
- process has died exit code 11 HOT 4
- How to get a trajectory format that can be used for evo tool ? HOT 1
- Intitialize in known geographic coordinates HOT 1
- fastlio+livox HOT 2
- make livox driver dependency optional HOT 1
- Memory leak HOT 15
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from fast_lio.