Trying to understand how the optimization parameters work for Lio-mapping

[ghost]

Hey. My question is are follows. What do the following parameters do in the indoor_test_config.yaml file in the lio_mapping package?

optimization options

run_optimization: 1

update_laser_imu: 1
gravity_fix: 1

plane_projection_factor: 0

imu_factor: 1
point_distance_factor: 1

prior_factor: 0

marginalization_factor: 1

odom_io: 6

pcl_viewer: 0

I am especially interested in odom_io as it seems to be affecting the trajectory of my odometry, but I am not sure how it affects it.

Thanks

[hyye]

Hi @SoguMax, thanks for your interest.
For some of the parameters you may refer to #20 (comment).
The odom_io is used to skip several frames. For example, odom_io: 2 means that it will take 1 point cloud from every 2 point clouds.

[ghost]

Thank you so much for your help.

…

On Tue, May 26, 2020 at 5:50 AM Haoyang Ye ***@***.***> wrote: Hi @SoguMax <https://github.com/SoguMax>, thanks for your interest. For some of the parameters you may refer to #20 (comment) <#20 (comment)>. The odom_io is used to skip several frames. For example, odom_io: 2 means that it will take 1 point cloud from every *2* point clouds. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#66 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APWXNLN5VNV4STPB6E4XD43RTM4A5ANCNFSM4NJLR6MA> .

-- [image: logo] *Samuel Ogunniyi* *R&D Engineer* +27 840 630434 <+27+840+630434> <http://za.linkedin.com/in/samuel-ogunniyi-56a1b377>

[ghost]

Hi, @hyye There seems to be a drift in the z direction when moving and capturing data through several floors in an indoor environment. What do you recommend? Moving the lidar and imu through a faster motion (The bag file is recorded with a slow walking pace) or just adjusting this odom_io parameter. Thank you

[gisbi-kim]

@SoguMax
similar issue.
I'm testing in an outdoor environment
also z drift occurs,
any advice.. ?

[hyye]

Hi @SoguMax and @kissb2, thanks for your interest. As I mentioned in #18 (comment), this drift is inevitable if the path is long and of no local loop. Adding a loop detection module, e.g., segmap, and optimize the pose graph may help.