Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

How can I limit memory usage? #6

Open
woshinidiesb opened this issue Dec 5, 2024 · 5 comments
Open

How can I limit memory usage? #6

woshinidiesb opened this issue Dec 5, 2024 · 5 comments

Comments

@woshinidiesb
Copy link

Hi, I'm here to provide feedback on the issues I encountered while using Fast Limo.

  1. During vehicle operation on bumpy roads, the vehicle's suspension design causes the radar to move abnormally, but Fast Limo did not consider this situation and will put incorrect point clouds into the map, affecting alignment.
  2. After enabling Fast Limo's local map mode, it began to malfunction. Adjusting the size parameter of the local map did not improve the situation. The map appeared with double images, and it began to drift.
    Do you have any good suggestions for these two issues?
@woshinidiesb
Copy link
Author

Uploading 截图 2024-12-05 13-49-41.png…

@woshinidiesb
Copy link
Author

The map appeared with double images that just like this

@fetty31
Copy link
Owner

fetty31 commented Dec 13, 2024

Hi @woshinidiesb,

I' m afraid I cannot see the image you shared. It would be nice if you could re-upload it.

For your first issue I don't know what can be done to improve its performance in this situation. The way I see it, if the bump is captured by the onboard IMU, FastLIMO should be able to compensate the points accordingly. For your second issue, as far as I am concerned, you are currently using the ros2-humble branch. In most of my projects I'm still working with Noetic, so the ROS2 branches are not extensively tested. Maybe this dynamic map feature is not well implemented in your working branch. I will check that when I can. Having said that, limiting the map size can have negative effects when the environment has few features in it (low texture) so make sure you are not limiting too much the size of the map. You can actually visualize the 3D box of the current map in Rviz in the topic fast_limo/map/bb .

@woshinidiesb
Copy link
Author

woshinidiesb commented Dec 13, 2024 via email

@fetty31
Copy link
Owner

fetty31 commented Dec 13, 2024

What dimensions of the local map are you trying? What RAM limitations do you have?

My data cat.bag can be sensitive to small map sizes because its actually a Formula Student driverless car in a completely empty parking lot. So in fact the algorithm is localizing with far away static objects like parked cars, some geometric of the parking structure and maybe some of the cones, hopefully. If the map is little, its impossible that the found features give useful information for SLAM.

If you are not moving, the map is still updated. However, as a voxelgrid is being applied, no much data is actually saved into the map. Also, you are seeing the same points over and over again so no new data is added. If this is not enough for you, more agressive filtering approaches can be performed to the pcl just before adding it to the map or maybe a fixed map size (max number of points) could developed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants