hku-mars / FAST_LIO

A computationally efficient and robust LiDAR-inertial odometry (LIO) package
GNU General Public License v2.0
2.69k stars 904 forks source link

point timestamp about VLP16 #55

Closed chengwei0427 closed 3 years ago

chengwei0427 commented 3 years ago

Hi,Dr Xu I have some problems about timestamp processing. added_pt.curvature = pl_orig.points[i].time / 1000.0; in preprocess.cpp and const double &pcl_end_time = pcl_beg_time + pcl_out.points.back().curvature / double(1000); in IMU_Precessing.hpp The point time stamp uint is microseconds in your code, and the cloud per frame timestamp is seconds, is that so? But in my vlp16 data, it looks like the point timestamp is in seconds, the VLP driver version is 1.6.0. 2021-08-12 10-47-14屏幕截图 What is your VLP driver version?

Thank you.

XW-HKU commented 3 years ago

You are right, the added_pt.curvature saves the offset time of each LiDAR point in ms, such that pl_orig.points[i].time's unit should be us. If your data's each LiDAR point's time is not us, you may consider modifying the FAST-LIO or your data.

I only test in the open datasets such as LIO-SAM's VLP-16 dataset, UTBM / NCLT / UrbanLoco 's HDL-32E datasets. Their point time all are in us.

chengwei0427 commented 3 years ago

OK , Thank you for your patient reply.