Commit Graph

22 Commits (eb96c91473764e41d86832741b46a76406822bc2)

Author SHA1 Message Date
Christoph Schütte 5896ead32e Introduce timeout for global constraints. (#517)
This PR introduces a new option that specifies the number of seconds
after which global matcher searches are performed if no recent global
constraints have been found between the submap's and the node's
trajectory.
2017-09-14 12:11:54 +02:00
Wolfgang Hess 22f41d8e37 Store rotational histogram as part of the node data. (#503)
The compressed point cloud is no longer stored as part of the
node data to reduce memory consumption.
2017-09-07 11:01:03 +02:00
Wolfgang Hess a7fe8bd2ab Add filtered point cloud to TrajectoryNode::Data for 2D. (#483) 2017-08-28 11:17:53 +02:00
jie 42d8a8f005 Add filtered point clouds to TrajectoryNode::Data for 3D. (#482) 2017-08-25 16:56:50 +02:00
Christoph Schütte 96b71e227f Add histograms for pose residuals (#478)
* WIP Add histograms for pose residuals

* Address reviewers comments

* Address more comments.
2017-08-25 11:39:56 +02:00
jie 2fd2e86351 Add fixed frame pose weight into optimization problem options. (#477) 2017-08-24 15:40:48 +02:00
Wolfgang Hess 4d11a226ff Expose low resolution matching scores. (#470)
This is needed in preparation of adding a low resolution score histogram.
2017-08-23 17:48:49 +02:00
zjwoody 2e53586818 Also check low resolution match for 3D loop closure. (#468)
Related to #369.
2017-08-22 17:29:20 +02:00
Damon Kohler f765e55ea9 Use multi-res scan matching for loop closures too. (#374)
Improves loop closure scan matcher refinement.
2017-06-30 09:30:09 +02:00
Wolfgang Hess 92f154f561 Also use fixed covariances in 3D SLAM. (#308)
This follows #307.
2017-06-01 15:04:07 +02:00
Wolfgang Hess a2e52f81cf Remove 2D scan matcher covariances. (#307)
This replaces the scaled covariances derived from the Ceres
scan matcher by directly configurable weights. Using covariances
did not provide the expected benefit, and replacing the scaling
matrix by two values will allow a faster evaluation of the cost
function in the future.
2017-05-31 11:56:32 +02:00
Wolfgang Hess b2175f0d93 Remove covariance from the 3D Ceres scan matcher. (#200)
Instead of passing around fake covariances generated by the scan matcher,
we introduce appropriate values where they are needed.
2017-02-08 15:22:28 +01:00
Wolfgang Hess 71c951b370 Remove the 'log_residual_histograms' option. (#166)
This option only existed in 2D. Also other refactoring to make the 2D
sparse pose graph optimization more similar to 3D.
2016-12-19 17:57:12 +01:00
Wolfgang Hess 5af133c0dd Improve the sparse pose graph optimization. (#165)
Removes the loss function for intra-submap constraints in 3D SLAM,
as it is already done in 2D. Also removes a duplicate call to the
solver in 2D, and code cleanup.
2016-12-19 17:02:05 +01:00
Wolfgang Hess 574a56bbbc Remove the UKF from local 2D SLAM. (#140)
Changes default configuration to log loop closure matches.
2016-11-23 12:37:55 +01:00
Wolfgang Hess 3e64a803a4 Consistently name weight parameter. (#116)
All parameters scaling optimization residuals to determine relative
importance are now consistently named.
2016-10-28 16:39:45 +02:00
Wolfgang Hess a5b81ff159 Allow continuing after running a "final" optimization. (#109)
And a tiny cleanup of the submap visualization query.
2016-10-27 15:25:57 +02:00
Wolfgang Hess d2962c004f Remove the 'also_match_to_new_submaps' option. (#29)
Instead, using a different constraint_builder.sampling_ratio is preferred.
2016-09-30 16:00:10 +02:00
Wolfgang Hess e526a7022f Remove covariance computation from branch-and-bound. (#27)
These covariances were only used in 2D and are only an estimate.
Following 3D, we change 2D to use the (local) covariance computed
using Ceres.
2016-09-23 14:27:19 +02:00
Wolfgang Hess 0796d50b00 Removes max_covariance_trace option. (#25)
This option is no longer really useful to predict outliers.
It is removed and other parameters scaled to minimize the change
in behavior. And related cleanup.
2016-09-22 11:35:47 +02:00
Wolfgang Hess 54a7faf294 Optimize only once per submap in 2D. (#22)
The setting was unintentionally set to optimize twice per submap which
harms parallelism.
2016-09-08 14:36:33 +02:00
Damon Kohler 269c28cb0c Moves the contents of the cartographer subdirectory up one level. 2016-08-03 12:48:29 +02:00