Estimate model coefficients using recursive least squares (RLS) algorithm
Estimators
Use the Recursive Least Squares Estimator block to estimate the parameters of a system that is linear in the parameters. Such a system has the following form:
y and H are known quantities that you provide to the block to estimate θ. The block estimates θ using the recursive least squares algorithm (see [1]).
For a given time step, t, y(t) and H(t) correspond to the Output and Regressors inports of the Recursive Least Squares Estimator block, respectively. θ(t) corresponds to the Parameters outport. y(t), must be a real-valued scalar. H and θ are real-valued vectors of length N, where N is the number of parameters to be estimated.
For example, suppose you want to estimate a scalar gain, θ, in the system y = h^{2}θ. Here, y is linear with respect to θ. You can use the Recursive Least Squares Estimator block to estimate θ. Specify y and h^{2} as inputs to the Output and Regressor inports.
Initial guess of the values of the parameters to be estimated, specified as one of the following options (each option can change the block dialog):
Block sample time, specified as -1 or a positive scalar.
The default value is -1. The block inherits its sample time based on the context of the block within the model.
Recursive estimation algorithm, specified as one of the following (each option can change the block dialog):
Forgetting Factor
—
(Default) Forgetting factor algorithm
If you select this option, you must specify the Forgetting Factor, λ, as a scalar in the (0 1] range. λ specifies the measurement window relevant for parameter estimation. Suppose the system remains approximately constant over T_{0} samples. You can choose λ such that:
$${T}_{0}=\frac{1}{1-\lambda}$$
Setting λ = 1 corresponds to "no forgetting" and estimating constant coefficients. Setting λ < 1 implies that past measurements are less significant for parameter estimation and can be "forgotten." Set λ < 1 to estimate time-varying coefficients. Typical choices of λ are in the [0.98 0.995] range.
The default value is 1.
Kalman Filter
—
Kalman filter algorithm
If you select this option, you must specify the Noise Covariance Matrix as one of the following:
Real nonnegative scalar, α — Covariance matrix is an N-by-N diagonal matrix, with α as the diagonal elements.
Vector of real nonnegative scalars, [α_{1},...,α_{N}] — Covariance matrix is an N-by-N diagonal matrix, with [α_{1},...,α_{N}] as the diagonal elements.
N-by-N symmetric positive semidefinite matrix.
N is the number of parameters to be estimated.
0 values in the noise covariance matrix correspond to estimating constant coefficients. Values larger than 0 correspond to time-varying parameters. Large values correspond to rapidly changing parameters.
The default value is 1.
Normalized Gradient
—
Normalized gradient adaptation algorithm
If you select this option, you must specify the following:
Adaptation Gain — Adaptation gain, γ, specified as a real positive scalar. γ is directly proportional to the relative information content in the measurements. That is, when your measurements are trustworthy, specify a large value for γ.
The default value is 1.
Normalization Bias — Bias in adaptation gain scaling, Bias, specified as a real nonnegative scalar. The normalized gradient algorithm scales the adaptation gain at each step by the square of the two-norm of the gradient vector. If the gradient is close to zero, this can cause jumps in the estimated parameters. Bias is the term introduced in the denominator to prevent these jumps. Increase Bias if you observe jumps in estimated parameters.
The default value is eps
.
Gradient
— Unnormalized
gradient adaptation algorithm
If you select this option, you must specify the Adaptation Gain, γ, as a real, positive scalar. γ is directly proportional to the relative information content in the measurements. That is, when your measurements are trustworthy, specify a large value for γ, and vice versa.
The default value is 1.
For more information about these algorithms, see Recursive Algorithms for Online Parameter Estimation.
Add Error outport to the block. Use this signal to validate the estimation.
For a given time step, t, the estimation error is calculated as:
$$e(t)=y(t)-yest(t).$$
Add Error outport.
(Default) Do not add Error outport.
Add Covariance outport to the block. Use this signal to examine parameter estimation uncertainty.
This option is available only when Estimation Method is
either Forgetting Factor
or Kalman
Filter
.
The software computes parameter covariance P
assuming
that the residuals, e(t), are
white noise, and the variance of these residuals is 1. The interpretation
of P
depends on the estimation method:
Forgetting Factor
— R_{2}/2
* P
is
approximately equal to the covariance matrix of the estimated parameters,
where R_{2} is the true variance
of the residuals.
Kalman Filter
— R_{2}* P
is
the covariance matrix of the estimated parameters, and R_{1} /R_{2} is
the covariance matrix of the parameter changes. Where, R_{1} is
the covariance matrix that you specify in Parameter Covariance
Matrix.
Add Covariance outport.
(Default) Do not add Covariance outport.
Add Enable inport to the block. Use this input signal to specify a control signal that enables or disables parameter estimation. The block estimates the parameter values for each time step that parameter estimation is enabled. If you disable parameter estimation at a given step, t, then the software does not update the parameters for that time step. Instead, the block outputs the last estimated parameter values. Use this option, for example, to disable parameter estimation when the system enters a mode where the parameter values do not vary with time.
Add Enable inport.
(Default) Do not add Enable inport.
Option to reset estimated parameters and parameter covariance matrix using specified initial values.
Suppose you reset the block at a time step, t. If the block is enabled at t, the software uses the initial parameter values specified in Initial Estimate to estimate the parameter values. In other words, at t, the block performs a parameter update using the initial estimate and the current values of the inports. The block outputs these updated parameter value estimates using the Parameters outport.
If the block is disabled at t and you reset the block, the block outputs the values specified in Initial Estimate.
Use this option, for example, when you reset the input because it did not excite the system as needed, resulting in poor estimation results.
Specify this option as one of the following:
None
— (Default)
Estimated parameters and covariance matrix values are not reset.
Rising
— Triggers
reset when the control signal rises from a negative or zero value
to a positive value. If the initial value is negative, rising to zero
triggers reset.
Falling
— Triggers
reset when the control signal falls from a positive or a zero value
to a negative value. If the initial value is positive, falling to
zero triggers reset.
Either
— Triggers
reset when the control signal is either rising or falling.
Level
— Triggers
reset in either of these cases:
Control signal is nonzero at the current time step
Control signal changes from nonzero at the previous time step to zero at the current time step
Level hold
— Triggers
reset when the control signal is nonzero at the current time step.
When you choose any option other than None
,
the software adds a Reset inport to the block. You provide the reset
control input signal to this inport.
Port Name | Port Type | Description |
---|---|---|
Regressors | In | H(t), specified as a real scalar or vector. |
Output | In | y(t), specified as a real scalar signal. |
Parameters | Out | θ(t), returned as a vector of the same length as H(t). |
Enable (Optional) | In | Control signal to enable parameter estimation, specified as a scalar. |
Reset (Optional) | In | Control signal to reset parameter estimation, specified as a scalar. |
InitialParameters (Optional) | In | Initial guess of the values of the parameters to be estimated, specified as a scalar or vector. |
InitialCovariance (Optional) | In | Initial covariance of parameters, specified as a real positive scalar, vector of real positive scalars, or symmetric positive-definite matrix. |
Error (Optional) | Out | Estimation error, returned as a scalar. |
Covariance (Optional) | Out | Covariance P of estimated parameters, returned
as an N-by-N matrix, where N is
the number of parameters. The software computes parameter covariance P assuming
that the residuals, e(t), are
white noise, and the variance of these residuals is 1. The interpretation
of P depends on the estimation method you specify
in Estimation Method:
This port is not available when Estimation
Method is either |
Double-precision floating point
Single-precision floating point
Note: The Regressors and Output inports must have matching data types. |
[1] Ljung, L. System Identification: Theory for the User. Upper Saddle River, NJ: Prentice-Hall PTR, 1999, pp. 363–369.