Exwayz software allows to perform complex trajectory optimization task that allow to take into account constraints from various nature.

odometry measures They are computed by ‣. These measures ensure the temporal consistency of the trajectory: consecutive trajectory points are relatively positioned by these measures.
loop closure observations They are computed by ‣. These observations add local constraints to the trajectory on areas visited multiple times and that might be offset due to the SLAM drift.
GNSS observations These are absolute position constraints provided by a global positioning system.

Understanding the loop closure optimization

The loop closure software searches for matches between trajectory indices by doing place recognition using the LiDAR 3D measures. Is goal is to identify “how should be the trajectory” on areas visited multiple times presenting drift.

figures (1).png

The output of this place recognition step is a set of loop closure observations between trajectory indices.

figures (2).png

Exwayz trajectory optimizer takes these measures into input and compute non-rigid deformation to the trajectory for integrating these observations and make the trajectory consistent.

figures.png

Understanding the georeferencing

The georeferencing takes in input odometry measures (from an arbitrary reference frame) and absolute position measures from a GNSS source.

Measure association

In order to use these in the same optimization process, the optimizer need to do a pairing between the odometry measures and absolute position errors, which is done using the timestamp of each measure.

<aside> 🙂 This is why we need an excellent time synchronization between the LiDAR data and the GNSS data

</aside>

figures (4).png

On the scheme above, a blues circles represent the -s option of the exwayz_georef software, which is the approximate confidence of each absolute measure. See exwayz_georef.exe for more info.

Now that we have pairs between odometry measures and absolute position measures, we can pre-align both trajectory.

figures (5).png

Most of odometry measures don’t lie in the incertitude ellipse of their absolute position match. This is because the GNSS system and the LiDAR don’t describe the trajectory of the same physical point. They actually don’t measure the position of the same point