- Introduce `RangeDataInserterInterface` as common interface for all range data inserters
- Introduce a minimal `GridInterface` as a base for a common interface for 2D and 3D grids
- Rename `RangeDataInserter2D` to `ProbabilityGridRangeDataInserter2D`
- Move grid generation logic from `Submap2D` to `ActiveSubmaps2D`
- Update proto and configuration structure to mirror the code structure - backwards compatibility is maintained
- Step towards [RFC 0019](https://github.com/googlecartographer/rfcs/blob/master/text/0019-probability-grid-and-submap2d-restructuring.md)
This introduces scheduling of tasks with dependencies to ThreadPool.
Contrary to work items, tasks are only queued for execution after all
their dependencies have completed.
This introduces the new class common::Task. A Task can have dependencies to other tasks, notify ThreadPoolInterface when all its dependencies are fulfilled, and can be executed in the background.
Since it is only the tests in cartographer/cloud that cannot run
concurrently with each other, use `tags = ["exclusive"]` to say that
they should not be run at the same time as other tests.
This works locally, not just in CI, and should speed up the CI job by
2-3 minutes.
- Base `ConstraintBuilder2D` and `FastCorrelativeScanMatcher2D` on correspondence cost function instead of probabilities
- No changes in the options of `ConstraintBuilder2D` and `FastCorrelativeScanMatcher2D`
- Add properties `min/max_correspondence_cost` to `Grid2D` and `proto::Grid2D`
- Provide backwards compatibility for `proto::Grid2D`
- Step towards [RFC 0019](https://github.com/googlecartographer/rfcs/blob/master/text/0019-probability-grid-and-submap2d-restructuring.md)
This is a first step towards generalizing submaps (see [ongoing RFC discussion](https://github.com/googlecartographer/rfcs/pull/30)).
Although this PR already introduces a correspondence cost grid, the grid is still used as probability grid.
Removes the confusing term `initial_pose` from global 2D optimization.
It makes sense in other areas like scan matching, but has a completely
different meaning there. Also matches the 3D equivalents better.
- `local_pose_2d` replaces `initial_pose` as the name for the
non-gravity-aligned, local 2D pose in the submap
- `global_pose_2d` replaces `pose` as the name for the
non-gravity-aligned, global 2D pose that is optimized
See the comment in the PR for how this improves readability.
gaschler
Christoph Schütte
Kevin Daun
Michael Grupp
Martin Schwörer
Alexander Belyaev
Holger Rapp
Rodrigo Queiro
Julius Kammerl