|
SUPPORT SITE |
|
|
|
SUPPORT SITE |
|
Path: Common/Estimation
--------------------------------------------------------------------------
Continuous discrete iterated extended Kalman Filter.
The state and covariance matrices are numerically integrated.
The model is of the form:
dx/dt = f(x,t,u)
y = h(x,t,u)
F(x,t,u) = df/dx
H(x,t) = dh/dx
x is the estimated state.
fName is a function of the form
[xDot, d.fData] = fName( x, t, d.fData )
hName is a function of the form
[z, dhdx] = hName( x, t, d.hData, d.meas.k )
If dhdx is empty, CDKF will linearize hName about x and t.
k are the indices to the measurements for which there is data.
Since version 9.
--------------------------------------------------------------------------
Form:
d = CDKF( d, fName, hName )
--------------------------------------------------------------------------
------
Inputs
------
d (:) Data structure
.time Time or jD
.year
.month
.day
.hour
.minute
.second
.timeLast Time of last call
.secFromStart Seconds from start
.x State
.p Covariance
.q Plant noise matrix
.k Gain matrix
.r Measurement noise matrix
.nIterations Number of iterations
.meas Measurements
.z New measurement(s)
.k Indices of new measurement(s)
.fData f function data
(user defined)
.hData h function data
(user defined)
fName (1,:) Dynamics function name
hName (1,:) Measurement function name
-------
Outputs
-------
d (:) Data structure
--------------------------------------------------------------------------
References: Gelb, A. Ed., Applied Optimal Estimation, MIT Press. p.188.
Table 6.1-1. Also, pp. 190-191.
--------------------------------------------------------------------------
Common: Time/Date2JD Math: Analysis/Jacobian
Back to the Common Module page