Parametric Solve Whiteboard

Parametric solving is currently under development, with initial linear or low dimensional PoC starting to work (4Q2019).

Next steps are:

• Consolidate Tree messages and message channels with non-parametric;
• (#467) Consolidate factor definitions with existing factor residual functions (i.e. drop "ParametricFactor" definitions);
• Calculate the covariance of each clique to ensure Gaussian belief propagation can be performed;
• Expand to manifolds and mixed manifolds;
• Integrate parametric solution alongside non-parametric solution through common API and defaults.

How to get Covariance Estimates

• https://onlinelibrary.wiley.com/doi/pdf/10.1002/9780470824566.app1
• https://julianlsolvers.github.io/Optim.jl/stable/#examples/generated/maxlikenlm/
• https://apps.dtic.mil/dtic/tr/fulltext/u2/a537233.pdf
• http://ais.informatik.uni-freiburg.de/publications/papers/tipaldiIROS07.pdf

Reference Issues or PRs

• https://github.com/JuliaRobotics/IncrementalInference.jl/pull/468
• https://github.com/JuliaRobotics/IncrementalInference.jl/issues/467