Richard Roberts
|
8ff5bf5c7c
|
Bug fix in BayesTree marginal, re-enabled joint and unit tests
|
2010-10-22 22:11:23 +00:00 |
Chris Beall
|
69c6d05ce1
|
config() changed to values() for consistency
|
2010-10-22 19:29:15 +00:00 |
Richard Roberts
|
f9e0ed07a3
|
Code cleanup: lean and mean VariableIndex (got rid of different types for different FG's, slots, and storage template), renamed Conditional.h and Factor.h to match class names ConditionalBase and FactorBase (not ideal names but prevents conflict with typedefs), added typedef for symbolic multifrontal solver.
|
2010-10-22 18:02:55 +00:00 |
Richard Roberts
|
1d94dacca0
|
make check passes with dense qr using lapack geqrf
|
2010-10-22 03:51:10 +00:00 |
Richard Roberts
|
812e3277ee
|
New linear solver interface, global inference functions removed.
|
2010-10-21 22:59:54 +00:00 |
Richard Roberts
|
f15fea202e
|
Change that may be reverted again soon - templated factor base type on key and make IndexFactor and IndexConditional for all current code. Did this to experiment with using symbols on the linear side in an efficient way but am putting this on hold for now because of complications. Will revisit and either fix or revert to int-only later in the week. Also moved some files to experimental and templating EliminationTree on FACTOR instead of FACTORGRAPH.
|
2010-10-19 21:31:13 +00:00 |
Chris Beall
|
4071d4f673
|
small change to make boost work properly
|
2010-10-18 21:44:38 +00:00 |
Richard Roberts
|
4a7b8bad27
|
--with-boost no longer required, automatically finding boost
|
2010-10-17 17:06:28 +00:00 |
Alex Cunningham
|
a9a066aec7
|
Removed hard constraints from gtsam library (with the exception of NonlinearEquality) and moved them to gtsam_experimental and MastSLAM
|
2010-10-17 03:56:05 +00:00 |
Kai Ni
|
bc36e653fd
|
depend on external CCOLAMD and DenseQR
|
2010-10-16 01:55:47 +00:00 |
Richard Roberts
|
130d9d2797
|
Made EliminationTree generic, with Symbolic and Gaussian unit tests
|
2010-10-15 15:53:36 +00:00 |
Kai Ni
|
24d499039f
|
prepend license information on all the codes
|
2010-10-14 04:54:38 +00:00 |
Kai Ni
|
5cbf67eeba
|
move colamd and spqr_mini out of gtsam
|
2010-10-14 02:41:08 +00:00 |
Richard Roberts
|
47b11749fe
|
Moved old elimination tree to gtsam_experimental
|
2010-10-13 22:07:52 +00:00 |
Richard Roberts
|
e545f59fb2
|
Inference::Marginal returns a factor graph
|
2010-10-13 20:41:26 +00:00 |
Frank Dellaert
|
b56b0fe3a4
|
Moved test to inference and added ETree prototype code
|
2010-10-12 05:19:51 +00:00 |
Frank Dellaert
|
fbdf9647f8
|
new test to drive elimination tree development
|
2010-10-12 00:15:26 +00:00 |
Frank Dellaert
|
057050fa9f
|
renamed varid_t to Index
|
2010-10-11 21:14:35 +00:00 |
Richard Roberts
|
96eb939749
|
Fixed computing marginals in BayesTree
|
2010-10-11 16:02:01 +00:00 |
Alex Cunningham
|
ccea5c79cb
|
unwhitened error in LinearApproxFactor now works
|
2010-10-11 04:47:57 +00:00 |
Alex Cunningham
|
c92026884b
|
moved and enabled testKey, LinearApproxFactor creation/linearization works
|
2010-10-11 04:30:19 +00:00 |
Alex Cunningham
|
c57c93a490
|
Re-enabled more constraints
|
2010-10-11 04:30:18 +00:00 |
Richard Roberts
|
45456aab8e
|
Fixed computing shortcuts in BayesTree
|
2010-10-10 03:10:03 +00:00 |
Richard Roberts
|
84d6b5be6a
|
Re-enabled computing marginals of a set of variables (Inference::Marginal(...))
|
2010-10-10 00:51:57 +00:00 |
Alex Cunningham
|
07bda5aa97
|
Changed linear config names to *Values, updated comments
|
2010-10-09 03:09:58 +00:00 |
Alex Cunningham
|
6002931e12
|
Renamed LieConfig and TupleConfig to use Values
|
2010-10-09 03:09:55 +00:00 |
Richard Roberts
|
1d52ff90a8
|
Merged simplelinear branch into trunk
|
2010-10-08 22:04:47 +00:00 |
Alex Cunningham
|
f6ada87314
|
Removed LDL, ConstraintOptimizer things, and FusionTupleConfig. Added *.valgrind run commands so that tests can be checked for memory errors. Removed some extraneous stuff in configure.ac
|
2010-10-02 23:38:01 +00:00 |
Alex Cunningham
|
5745226452
|
split LinearApproxFactor into an implementation header file
|
2010-09-22 15:34:03 +00:00 |
Alex Cunningham
|
b09b7fffbb
|
Removed extraneous template parameter for LinearApproxFactor
|
2010-09-22 13:29:29 +00:00 |
Frank Dellaert
|
3f937fdba4
|
changed include order to be more logical
|
2010-09-20 21:02:32 +00:00 |
Richard Roberts
|
55dedc97e3
|
Fixed include paths so that making individual unit tests works again
|
2010-09-11 16:54:50 +00:00 |
Alex Cunningham
|
815d892806
|
Added TEST_UNSAFE to avoid exception checking in normal tests, as well as an assert_inequal() evaluation in TestableAssertions.h.
|
2010-08-30 20:54:12 +00:00 |
Alex Cunningham
|
f594ebf562
|
Reworked testSerialization so that it works for geometry and simple planar objects. Currently only tested in boost 1.40, probably doesn't work under boost 1.44
|
2010-08-30 19:16:26 +00:00 |
Alex Cunningham
|
a461680a2b
|
Removed extraneous headers that were causing problems related to boost serialization and added testSerialization with conditional compiling (--enable-serialization). Currently, the test is all commented out, and it should not affect those not using the flag.
|
2010-08-27 15:31:20 +00:00 |
Alex Cunningham
|
23a30f8475
|
Removed remaining global lie functions on lie objects and configs, switched the Lie base class to a simple concept check function, fixed build script for examples. ISAM2 and MastSLAM verified as compiling.
|
2010-08-26 19:55:40 +00:00 |
Alex Cunningham
|
eed13e48d2
|
Switched to using LieVectors for vector types. Still some problems with numericalDerivative that have been worked around, but other functionality is intact.
|
2010-08-24 17:26:56 +00:00 |
Alex Cunningham
|
77eda5ab8c
|
In all nonlinear factors and configs, it is only necessary to specify a key with a typedef Value_t for the associated value. This has been removed from factor template definitions, as it is too easy to use the wrong value. Also, no more string keys or linear symbol keys in nonlinear systems. Updated MastSLAM to work, and ISAM2 works without change
|
2010-08-23 19:44:17 +00:00 |
Alex Cunningham
|
5350f368cd
|
Removing binary
|
2010-08-23 03:44:25 +00:00 |
Alex Cunningham
|
e0da1971fa
|
fixed ROS packages due to gtsam changes
|
2010-08-23 00:19:57 +00:00 |
Alex Cunningham
|
458cc52fff
|
Switched geometry to use primarily member functions for calculations with optional derivatives, so there are no more optional derivative functions. Also split Vector specializations for Lie into gtsam/base/LieVector.h which will later change into a real wrapper function. Specialized numericalDerivative to allow for functions to return doubles. Projects ISAM2, MastSLAM updated.
|
2010-08-22 21:45:53 +00:00 |
Richard Roberts
|
c2d46ce8b3
|
Script to time the overhead of breaking up a problem into many factors (mimics a linear latent-variable model)
|
2010-08-21 17:54:59 +00:00 |
Alex Cunningham
|
256697d345
|
Moved transform_[to|from] and [un]rotate to be static functions in classes
|
2010-08-20 21:47:30 +00:00 |
Richard Roberts
|
f1132359d4
|
Installing headers in subdirectories and converted include directives to match the subdirectory structure (also see email to frankcvs)
|
2010-08-19 17:23:19 +00:00 |
Alex Cunningham
|
30c3f46b52
|
FusionTupleConfigs work correctly for optimization
|
2010-08-16 19:19:50 +00:00 |
Alex Cunningham
|
977155d133
|
Added simpler interfaces for FusionTupleConfig for 1,2, and 3 variable types
|
2010-08-15 21:35:03 +00:00 |
Alex Cunningham
|
a2fecf51c9
|
FusionTupleConfig is now a full config
|
2010-08-14 21:15:45 +00:00 |
Alex Cunningham
|
f3ccfaf100
|
Added dummy initialization of LieConfigs from arbitrary other LieConfigs, FusionTupleConfigs can now be created from arbitrary other FusionConfigs.
|
2010-08-13 19:51:04 +00:00 |
Alex Cunningham
|
f6dbee41d2
|
More of FusionTupleConfig works
|
2010-08-13 14:55:26 +00:00 |
Alex Cunningham
|
c2a83759cb
|
Added experimental replacement for TupleConfig using boost.fusion - this should eventually make it possible to make nonlinear factors interchangable regardless of config type.
|
2010-08-12 15:23:03 +00:00 |