search

sisl / GaussianFilters.jl

Julia Package for discrete-time linear Gaussian parametric filtering systems, namely KF, EKF, UKF, GM-PHD
MIT License
44 stars 11 forks source link

Additional filter implementations (DD2) #36

Open radioflash opened 3 years ago

radioflash commented 3 years ago

Some time ago, I played around with the DD2 Kalman filter--see "New developments in state estimation for nonlinear systems" by Nørgaard, Poulsen, Ravn in Automatica 36 (11), 1627-1638 (doi:10.1016/s0005-1098(00)00089-3; it's on scihub).

From what I can tell, this is kinda similar to the UKF filter (a nonlinear extension of classic Kalman filtering, not relying on explicitly supplied derivatives). It is based on Stirling interpolation, and claims better performance (specifically covariance estimation) than the UKF.

There is some Matlab reference code, but ready-to-use library implementations are not common (that I found).

I would base any work on the current UKF implementation (because some implementation details seem quite similar to me). I also have some real world data to benchmark them against each other.

Would it make sense to add an implementation to this package? Does anyone have any input/opinion on this (interesting/redundant/pointless/belongs elsewhere)?

zsunberg commented 3 years ago

Have not thought through in detail, but another option would be to create a new package that depends on this one with the advanced DD2 filter.