HI, I use the nanoicp to location, but it cann`t work about direct_lidar_odometry HOT 10 CLOSED

I found that the translation grows fast，it seems strangeness

from direct_lidar_odometry.

Hi @tust13018211 -- are you using NanoGICP as part of DLO or outside of this package? Thanks.

from direct_lidar_odometry.

I try to use in dlo and outside , all the result is not good

from direct_lidar_odometry.

from direct_lidar_odometry.

Are covariances and/or kdtrees of both clouds being properly set in your code?

from direct_lidar_odometry.

is there any pronlem above ?

from direct_lidar_odometry.

at first, I use the
m_iGicpScanToMap_.setInputTarget(map);
m_iGicpScanToMap_.calculateTargetCovariances();
to built target cloude Covariances and kdtree, and then
m_iGicpScanToMap_.setInputSource(m_pCurrentScan_);
m_iGicpScanToMap_.calculateSourceCovariance
use this to built source cloude Covariances and kdtree, at last,
m_iGicpScanToMap_.align(*aligned, init_pose.matrix.cast());

is there any pronlem above ?

from direct_lidar_odometry.

Hmm looks fine to me. Just to clarify: does DLO work on your end (without any modifications)? And in your code above, switching out NanoGICP with PCL's GICP works? The package should work as a replacement for PCL's GICP (sans some of the computational benefits from data structure recycling). A bit hard to help debug without more information.

from direct_lidar_odometry.

Thaks your reply, I just want to test nanogicp for location with a map, I adjust the parameter, it seems work. the result in the picture, the result seems not very good

from direct_lidar_odometry.

Oh I see. Unfortunately we have not tested DLO (or NanoGICP) for localization with a provided map (as mentioned here and here) and assumes no prior knowledge of the environment, so your results may vary.

from direct_lidar_odometry.