rcppkalman
Kalman filtering via RcppArmadillo -- based on a R and C++ port of the EKF/UKF toolbox for Matlab
Why?
R has a number of existing Kalman filter packages which are all very noteworthy in their own right. A comparison was provided by this JSS paper from 2011.
Yet I had a need for something both simple and fast at the C++ level.
The EKF/UKF toolbox for Matlab proved to be a wonderful source of excellent code that was well documented (see [this 130 page pdf manual]http://becs.aalto.fi/en/research/bayes/ekfukf/documentation.pdf()), under a suitable license and covering both simple examples as well as promising extensions.
Demos
Static Sine Signal And Noisy Measurement
This example is not described in the pdf manual, but included as demo kf_sine_demo.m within the EKF/UKF sources. A signal is provided via a sine wave plus random noise, and a linear Kalman Filter is used to smooth and filter the series. Our variant demo/kf_sine_demo.R reproduces the demo via the following chart
Continuous Wiener-Process Acceleration
This demo is described in detail in Section 2.2.4 on pages 11 to 15 of the EKF/UKF Documentation; the animation is part of the corresponding Matlab code in kf_cwpa_demo.m. We show the two final charts which provide animations of the smoothing and filtering in our version demo/kf_cwpa_demo.R:
Status
Working, but still far from complete. We currently support two demo scripts based on linear smoothers and filters. Additional functions should get added over time.
Author
The EKF/UKF Toolbox for Matlab was written by Simo Särkkä, Jouni Hartikainen, and Arno Solin.
Dirk Eddelbuettel is writing and maintaing this package by porting it to R and C++ via Rcpp and particularly RcppArmadillo.
License
This package is released under GNU General Public License, Version 2 or later. EKF/UKF itself (which is included) is released under the GNU General Public License, Versions 2 and 3.