Kalman filtering framework: ekfTransform(f, J_x, mu, P, Q, curTime, isLinearNoise, J_v) - File Exchange

Code covered by the BSD License

Highlights from
Kalman filtering framework

DoubleWell
HeatConduction simple model of 1-dimensional heat conduction
Reentry from Julier and Uhlmann's UKF article
ekfTransform(f, J_x, mu, P, Q...
sdchol(A) SDCHOL Cholesky-like decomposition for positive semidefinite matrices
symmetricSigmaPoints(n, W_mea... input:
unscentedTransform(f, mu, cho...
BLUEFilter BLUEFILTER Filter based on Best Linear Unbiased Estimator (BLUE)
BaseContinuousSystemModel BASECONTINUOUSSYSTEMMODEL Abstract base class for a continuous time system
BaseDiscreteSystemModel BASEDISCRETESYSTEMMODEL Abstract base class for discrete system models
BaseObserverModel BASEOBSERVERMODEL Abstract base class for observer models
BaseSystemModel BASESYSTEMMODEL Abstract base class for system models
EKFContinuousSystemModel EKFCONTINUOUSSYSTEMMODEL Class modeling a continuous time system with EKF predictions
EKFObserverModel EKFOBSERVERMODEL Observer model based on Extended Kalman Filter
EKFSystemModel EKFSYSTEMMODEL Class modeling a discrete time system with EKF predictions
LinearContinuousSystemModel LINEARCONTINUOUSSYSTEMMODEL Class modeling a linear continuous time system
LinearObserverModel LINEAROBSERVERMODEL Linear observer model
LinearSystemModel LINEARSYSTEMMODEL Class modeling a discrete system with linear evolution
UnscentedContinuousSystemModel UNSCENTEDCONTINUOUSSYSTEMMODEL Class modeling a continuous time system with EKF predictions
UnscentedObserverModel UNSCENTEDOBSERVERMODEL Observer model based on Unscented Kalman Filter
UnscentedSystemModel UNSCENTEDSYSTEMMODEL Class modeling a discrete time system with UKF predictions
View all files

from Kalman filtering framework by David Ogilvie
Object framework for filtering using Kalman filter, EKF, or UKF.

ekfTransform(f, J_x, mu, P, Q, curTime, isLinearNoise, J_v)

function [y S C] = ekfTransform(f, J_x, mu, P, Q, curTime, isLinearNoise, J_v)
%function [y S C] = ekfTransform(this, predMean, predCov, meas, curTime)
        % TODO: allow for multiple steps
    if nargin <= 6
        isLinearNoise = true;
    end

    H = J_x(mu, curTime);
    
    if isLinearNoise
        y = f(mu, curTime);
        S = H * P * H' + Q;
    else
        zeroNoise = zeros(size(this.Q,1), 1);
        y = f(mu, t, zeroNoise);
        M = J_v(mu, curTime);
        S = H * P * H' + M * Q * M';        
    end
    
    C = P * H';
end

Highlights from Kalman filtering framework

Highlights from
Kalman filtering framework