[IEEE ICRA'23] A new lightweight LiDAR-inertial odometry algorithm with a novel coarse-to-fine approach in constructing continuous-time trajectories for precise motion correction.
In the geometric observer paper ("A Contracting Hierarchical Observer for Pose-Inertial Fusion"), the state update is described by the following equations:
From these derivative equations, how to you get yours?
I understand your update comes from a first order derivative approximation, where the correction is applied a posteriori on the latest state, using the total integration time Dt:
In the geometric observer paper ("A Contracting Hierarchical Observer for Pose-Inertial Fusion"), the state update is described by the following equations:
From these derivative equations, how to you get yours?
I understand your update comes from a first order derivative approximation, where the correction is applied a posteriori on the latest state, using the total integration time
Dt
:where
s_imu
is the state computed from imu integration, based on the following equations:and
s_err'
are the "correction" part of the geometric obserser equations:The correction factors looks the same on both papers for q, bw, and p. However for v and ba, I think your implementation is missing a few factors:
Why is that? are these factors experimentaly negligible?