Path: Common/Estimation
% Rauch-Tung-Striebel smoothing for an Unscented Kalman Filter.
This is for a non-augmented, i.e. additive Gaussian noise form.
This assumes a model of the form:
x = f(x,q)
y = h(x,r)
You pass the data structure that was already generated by
an Unscented Kalman Filter and the results of the predict and update
steps.
--------------------------------------------------------------------------
Form:
d = UKFRTSS( m, p, d )
--------------------------------------------------------------------------
------
Inputs
------
m (n,:) Means
p (n,n,:) Covariance
d (1,1) UKF data structure
.m (n,1) Mean
.p (n,n) Covariance
.q (n,n) State noise
.wM (1,2n+1) Model weights
.w (2n+1,2n+1) Weight matrix
.f (1,:) Name of right hand side
.fData (1,1) Data structure with data for f
.dT (1,1) Time step
.t (1,1) Time
-------
Outputs
-------
d (1,1) UKF data structure
.mS (:,1) Mean smoothed
.pS (:,:) Covariance smoothed
--------------------------------------------------------------------------
Math: Integration/RK4 Math: Linear/DupVect
Back to the Common Module page