new partial correlation test

[dehann]

xref JuliaRobotics/IncrementalInference.jl#1206

replace #422

[codecov]

Codecov Report

Merging #434 (f94ba67) into master (6bbae00) will increase coverage by 0.37%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #434      +/-   ##
==========================================
+ Coverage   25.05%   25.43%   +0.37%     
==========================================
  Files          45       45              
  Lines        1724     1718       -6     
==========================================
+ Hits          432      437       +5     
+ Misses       1292     1281      -11     

Impacted Files	Coverage Δ
src/factors/Range2D.jl	50.00% <ø> (+28.94%)	⬆️
src/FactorGraphAnalysisTools.jl	33.33% <0.00%> (+1.19%)	⬆️
src/variables/Local_Manifold_Workaround.jl	50.00% <0.00%> (+1.72%)	⬆️
src/factors/Pose2Point2.jl	58.82% <0.00%> (+3.26%)	⬆️
src/factors/Point2D.jl	64.00% <0.00%> (+4.74%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6bbae00...f94ba67. Read the comment docs.

[dehann]

FIX in IncrementalInference v0.23.1

Correct Result For New Test

This is the correct result for the new test. Broad odometry error from first to second pose implies large uncertainty in second pose. The overlap in range uncertainty from both first and second pose result in two slanted (but narrower) modes around the y-axis for landmark L1:

Bad Result Example

This was unfortunately the nominal (but somewhat incorrect) result during the time after closing IIF 467 and before IIF v0.23.1.

---

This fix and new test will likely improve the result wrt to partials in

• Problems using multihypo on chains of relative measurements #378 and Problems using Pose2Point2BearingRange in larger examples #380 (cc @lemauee), and
• MWE was identified while working with @doublestrong

Note that work on known issues with partials and the Manifolds refactoring is ongoing. The recently closed #244 has been a big step towards more upcoming changes in this regard.