kalman

Design Kalman filter for state estimation.

Syntax

  • [kalmf, L, P, M, Z] = kalman(sys, Q, R, N)
  • [kalmf, L, P, M, Z] = kalman(sys, Q, R, N, sensors, known)

Input argument

  • sys - Plant model with process noise: state-space model.
  • Q - Process noise covariance: scalar or matrix.
  • R - Measurement noise covariance: scalar or matrix.
  • N - Noise cross covariance: scalar or matrix.
  • sensors - Measured outputs of sys: vector.
  • known - Known inputs of sys: vector.

Output argument

  • kalmf - Kalman estimator: state-space model
  • L - Filter gains: matrix
  • P - Steady-state error covariances: matrix
  • M - Innovation gains of state estimators: matrix
  • Z - Steady-state error covariances: matrix

Description

[kalmf, L, P] = kalman(sys, Q, R, N) generates a Kalman filter using the provided plant model sys and noise covariance matrices Q, R, and N.

The function calculates a Kalman filter suitable for application in a Kalman estimator, as depicted in the following diagram.

Example

A = [11.269 -0.4940 1.129; 1.0000 0 0;0 1.0000 0]; B = [-0.3832; 0.5919; 0.5191]; C = [1 0 0]; sys = ss(A,[B, B], C, 0); Q = 1; R = 1; [kEst, l, p, m, z] = kalman(sys, Q, R, [])

See also

care, dare.

History

VersionDescription
1.0.0initial version

Author

Allan CORNET