search

hku-mars / LiDAR_IMU_Init

[IROS2022] Robust Real-time LiDAR-inertial Initialization Method.
GNU General Public License v2.0
857 stars 141 forks source link

Question about implementation of `Forward_propagation_without_imu` function #17

Closed Taeyoung96 closed 2 years ago

Taeyoung96 commented 2 years ago

@zfc-zfc @Joanna-HE
Thanks for your wonderful works!

I want to apply the initialization method presented in LiDAR_IMU_Init directly to FAST-LIO algorithm.

But I was confused because the implementation is a little bit different. :cry:

To express the state vector, LiDAR_IMU_Init used struct StatesGroup but FAST-LIO used IKFoM_toolkit.

I'm not familiar with IKFoM, so I don't have an idea to implement ImuProcess::Forward_propagation_without_imu() function to FAST-LIO.

Would you mind giving me some tips?

In LiDAR_IMU_Init, StateGroup &state_inout and in FAST-LIO esekform::esekf<state_ikfom, 12, input_ikfom> &kf_state represent the same state vector?

What does the variable in kf_state correspond to the .cov of StateGroup?

I'd appreciate your help. :bow:

Thanks,

zfc-zfc commented 2 years ago

Hi. LiDAR_IMU_Init contains not only a full initialization module, but also the whole FAST-LIO2 (directly implemented by Eigen, not ikfom, but they are the same actually). In LiDAR_IMU_Init, it will first do initialization. After initialization finished, it will switch into FAST-LIO2 to refine extrinsic transformation of LiDAR and IMU. So, you don't need to apply initialization module to FAST-LIO algorithm because that has already been finished by LiDAR_IMU_Init.

Taeyoung96 commented 2 years ago

@zfc-zfc
I understand it! Thanks for your reply :)