Return measurement residual and residual covariance when using extended or unscented Kalman filter
residual command returns the difference between the
actual and predicted measurements for
unscentedKalmanFilter objects. Viewing the residual provides a way for
you to validate the performance of the filter. Residuals, also known as
innovations, quantify the prediction error and drive the
correction step in the extended and unscented Kalman filter update sequence. When using
predict (System Identification Toolbox) to update the estimated Kalman filter state, use the
residual command immediately before using the
returns the residual
Residual between a measurement
y and a predicted measurement produced by the Kalman filter
obj. The function also returns the covariance of the
obj using the
unscentedKalmanFilter commands. You
specify the state transition function f and measurement function
h of your nonlinear system in
State property of the object stores the latest estimated
state value. At each time step, you use
predict together to update the state
x. The residual s is the difference
between the actual and predicted measurements for the time step, and is expressed as
s = y -
h(x). The covariance of the residual
S is the sum R +
RP, where R is
the measurement noise matrix set by the
property of the filter and RP is the state
covariance matrix projected onto the measurement space.
Use this syntax if the measurement function h that you
obj.MeasurementFcn has one of the following forms:
y(k) = h(x(k))for additive measurement noise
y(k) = h(x(k),v(k))for nonadditive measurement noise
v(k) are the measured output, states, and measurement noise
of the system at time step
k. The only inputs to
h are the states and measurement noise.
Use this syntax if the measurement function h has one of the following forms:
y(k) = h(x(k),Um1,...,Umn)for additive measurement noise
y(k) = h(x(k),v(k),Um1,...,Umn)for nonadditive measurement noise
Estimate States Online Using Extended Kalman Filter
Estimate the states of a van der Pol oscillator using an extended Kalman filter algorithm and measured output data. The oscillator has two states and one output.
Create an extended Kalman filter object for the oscillator. Use previously written and saved state transition and measurement functions,
vdpMeasurementFcn.m. These functions describe a discrete-approximation to a van der Pol oscillator with the nonlinearity parameter equal to 1. The functions assume additive process and measurement noise in the system. Specify the initial state values for the two states as [1;0]. This is the guess for the state value at initial time
k, based on knowledge of system outputs until time
obj = extendedKalmanFilter(@vdpStateFcn,@vdpMeasurementFcn,[1;0]);
Load the measured output data
y from the oscillator. In this example, use simulated static data for illustration. The data is stored in the
load vdp_data.mat y
Specify the process noise and measurement noise covariances of the oscillator.
obj.ProcessNoise = 0.01; obj.MeasurementNoise = 0.16;
Initialize arrays to capture results of the estimation.
residBuf = ; xcorBuf = ; xpredBuf = ;
Implement the extended Kalman filter algorithm to estimate the states of the oscillator by using the
predict commands. You first correct using measurements at time
k to get . Then, you predict the state value at the next time step using , the state estimate at time step
k that is estimated using measurements until time
To simulate real-time data measurements, use the measured data one time step at a time. Compute the residual between the predicted and actual measurement to assess how well the filter is performing and converging. Computing the residual is an optional step. When you use
residual, place the command immediately before the
correct command. If the prediction matches the measurement, the residual is zero.
After you perform the real-time commands for the time step, buffer the results so that you can plot them after the run is complete.
for k = 1:size(y) [Residual,ResidualCovariance] = residual(obj,y(k)); [CorrectedState,CorrectedStateCovariance] = correct(obj,y(k)); [PredictedState,PredictedStateCovariance] = predict(obj); residBuf(k,:) = Residual; xcorBuf(k,:) = CorrectedState'; xpredBuf(k,:) = PredictedState'; end
When you use the
obj.StateCovariance are updated with the corrected state and state estimation error covariance values for time step
CorrectedStateCovariance. When you use the
obj.StateCovariance are updated with the predicted values for time step
PredictedStateCovariance. When you use the
residual command, you do not modify any
In this example, you used
predict because the initial state value was , a guess for the state value at initial time
k based on system outputs until time
k-1. If your initial state value is , the value at previous time
k-1 based on measurements until
k-1, then use the
predict command first. For more information about the order of using
correct, see Using predict and correct Commands.
Plot the estimated states, using the postcorrection values.
plot(xcorBuf(:,1), xcorBuf(:,2)) title('Estimated States')
Plot the actual measurement, the corrected estimated measurement, and the residual. For the measurement function in
vdpMeasurementFcn, the measurement is the first state.
M = [y,xcorBuf(:,1),residBuf]; plot(M) grid on title('Actual and Estimated Measurements, Residual') legend('Measured','Estimated','Residual')
The estimate tracks the measurement closely. After the initial transient, the residual remains relatively small throughout the run.
Specify State Transition and Measurement Functions with Additional Inputs
Consider a nonlinear system with input
u whose state
x and measurement
y evolve according to the following state transition and measurement equations:
The process noise
w of the system is additive while the measurement noise
v is nonadditive.
Create the state transition function and measurement function for the system. Specify the functions with an additional input
f = @(x,u)(sqrt(x+u)); h = @(x,v,u)(x+2*u+v^2);
h are function handles to the anonymous functions that store the state transition and measurement functions, respectively. In the measurement function, because the measurement noise is nonadditive,
v is also specified as an input. Note that
v is specified as an input before the additional input
Create an extended Kalman filter object for estimating the state of the nonlinear system using the specified functions. Specify the initial value of the state as 1 and the measurement noise as nonadditive.
obj = extendedKalmanFilter(f,h,1,'HasAdditiveMeasurementNoise',false);
Specify the measurement noise covariance.
obj.MeasurementNoise = 0.01;
You can now estimate the state of the system using the
correct commands. You pass the values of
correct, which in turn pass them to the state transition and measurement functions, respectively.
Correct the state estimate with measurement
0.8 and input
0.2 at time step
Predict the state at the next time step, given
Retrieve the error, or residual, between the prediction and the measurement.
[Residual, ResidualCovariance] = residual(obj,0.8,0.2);
obj — Extended or unscented Kalman filter
extendedKalmanFilter object |
y — Measured system output
Measured system output at the current time step, specified as an N-element vector, where N is the number of measurements.
Um1,...,Umn — Additional input arguments to measurement function
input arguments of any type
Additional input arguments to the measurement function of the system,
specified as input arguments of any type. The measurement function
h is specified in the
MeasurementLikelihoodFcn property of
obj. If the function requires input arguments in
addition to the state and measurement noise values, you specify these inputs
residual command syntax. The
residual command passes these inputs to the
measurement or the measurement likelihood function to calculate estimated
outputs. You can specify multiple arguments.
For instance, suppose that your measurement or measurement likelihood
function calculates the estimated system output
u and current time
addition to the state
Um2 terms are therefore
k. These inputs result in
the estimated output
y(k) = h(x(k),u(k),k)
Before you perform online state estimation correction at time step
k, specify these additional inputs in the
residual command syntax:
[Residual,ResidualCovariance] = residual(obj,y,u(k),k);
For an example showing how to use additional input arguments, see Specify State Transition and Measurement Functions with Additional Inputs.
Residual — Residual between current and predicted measurement
scalar | vector
Residual between current and predicted measurement, returned as a:
Scalar for a single-output system
Vector of size N for a multiple-output system, where N is the number of measured outputs
ResidualCovariance — Residual covariance
Residual covariance, returned as an N-by-N matrix where N is the number of measured outputs.
Introduced in R2019b