search

TixiaoShan / LIO-SAM

LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping
BSD 3-Clause "New" or "Revised" License
3.34k stars 1.24k forks source link

Odometry problem with some Kitti sequences #371

Open BurhanMuhyiddin opened 2 years ago

BurhanMuhyiddin commented 2 years ago

Hi. First of all, I would like to appreciate your excellent work. Thank you making it open source and sharing it. I have tested the algorithm with several kitti sequences. The interesting thing is that it works with some of the kitti sequences good, while with other sequences it gives very inaccurate results. Namely, the base_link starts to jump around and so, the results are awful. I attached picture below as an example. Screenshot from 2022-09-03 13-22-20

So, I started to debug the code. After a bit of debugging I think I found the problem and even I partially solved it, but I am sure that it is not correct solution and also, it doesn't completely solves the problem.

  1. Problem is not external IMU parameters, because it is the same for every Kitti sequences (I checked with calibration files also)
  2. Problem is not with bag file's IMU or Lidar datas, because I have checked them. Also, they work with Fast-LIO
  3. Problem is with LIDAR Odometry. As it is not correct, it directly affect to IMU odometry and base_link starts to jump around. After a bit of debugging I have seen it is related from scan2MapOptimization. Namely, during optimization, R and T is not converging. I mean, it converges, but for a few frames, so LIDAR odometry becomes non-accurate. I did some experience on that. I have increased convergence threshold from 0.05 to 0.5 (deltaR and deltaT) and also, maximum iteration number (iterCount) to 50. These parameters are in LMOptimization method. The results became better. Now the base doesn't jump around in a crazy way. While it is surely far more better result than previous case, it is still not that correct result. So, please can you provide me a solution for that problem? What should I do? How can I make optimization process better. Is it related because of not being able to match the features? Thank you very much.
stale[bot] commented 1 year ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

wh22920 commented 1 year ago

Hi，I also encountered the same problem, with the trajectory experiencing the same drift. Have you solved this problem and can you provide a solution