This changes which submaps we select to attempt loop closing.
The subsampling of candidates is changing from randomly sampling
submap and node pairs to per-submap sampling. This enforces a
better distribution of loop closure attempts across the submaps.
This sampling achieves a much better performance which indicates
that the approach used before was sub-optimal.
Signed-off-by: Wolfgang Hess <whess@lyft.com>
This is to make it possible to use the proto definitions
from Python where the directory structure otherwise
prevents them from being imported in the usual way.
Signed-off-by: Wolfgang Hess <whess@lyft.com>
The histogram of a submap is now stored in the submap (class and proto) itself. This change allows to accumulate the histogram of a submap in local SLAM by adding up the histogram of each new scan.
The main advantage is that the background thread doesn't have to loop over all `TrajectoryNode`s of a finished submap to compute the submap histogram for the `RotationalScanMatcher`. Instead this chunk of work is moved to the local SLAM thread but is split up into a few computations for each new scan. When running localization, the histogram of a submap can just be read from a map pbstream and does not have to be computed from the nodes.
In summary:
- This change improved the CPU time of offline SLAM by ~7%.
- Increases the readability of the code and performance of the background thread. (see `PoseGraph3D::ComputeConstraint`)
- No negative performance impacts on accuracy or finding loop-closures
However:
- With this change to the submap proto, old maps (pbstreams) are no longer supported and need to be re-created by running offline slam
The uplink server only receives the grid content of a submap after
that submap is finished for efficiency. Therefore, constraint
searches against that submap need to be skipped.
Also add checks to avoid this in the future.
FIXES=#1360
Wolfgang Hess
Michael Grupp
Martin Schwörer
Alexander Belyaev
gaschler
Arno Fleck
Sebastian Klose
Christoph Schütte