search

RozDavid / LOL

LOL: Lidar-only Odometry and Localization in 3D point cloud maps
353 stars 95 forks source link

Spikes on the vehicle path #11

Closed WilliamWoo45 closed 3 years ago

WilliamWoo45 commented 3 years ago

Hi @RozDavid,

I've successfully run the LOL on the KITTI dataset, and obtained the correct matching between the input point cloud & the map. However, I noticed there are some spikes along the vehicle path (shown as below), which is quite weird and not normal.

LOL_path_problem1

LOL_path_problem2

LOL_path_problem3

Any possible reason for this?

Thanks and Best Regards, William

RozDavid commented 3 years ago

Hey @WilliamWoo45,

Could you share what you have in your bag files? The point of the LOL algorithm is to run Odometry and Localization on the same streaming Lidar scans. If you broadcast the vehicle transforms from the bag on the /tf topic, strange behaviours like this could happen, also there is no point for using LOL for known vehicle trajectory.

For using LOL the correct way, one should have bag files with only the static tf transforms topic and the lidar scan topic - in your case the easiest solution could be to record a new bag file from the original Kitti dataset, where you drop the unnecessary topics.

WilliamWoo45 commented 3 years ago

Hi @RozDavid,

Thanks for your prompt response and detailed clarification. The method of recording a new rosbag works.

Best Regards, William