ADCS.satellite_hardware.errors.bias module

class ADCS.satellite_hardware.errors.bias.Bias(bias=array([0.]), std_bias=array([0.]), bounds=(array([-inf]), array([inf])))

Bases: object

Represents actuator bias modeled as a bounded random walk.

Parameters:

• bias (float, optional) – Initial bias value \(b_0\) (default 0).

• std_bias (float, optional) – Standard deviation rate \(\sigma_b\) controlling random-walk diffusion (default 0).

• bounds ((float, float), optional) – Clipping range \([b_{\min}, b_{\max}]\) applied after updates.

broadcast_to(length, *, owner='sensor')

Return this model sized for a length-output sensor.

A scalar model is repeated per axis, a model that already has length entries is copied, and any other size is a configuration error reported here, where the owning sensor is known, instead of surfacing later as a covariance-shape error inside the filter. Subclasses with their own parameterisation are copied untouched.

Parameters:

• length (int)

• owner (str)

Return type:

Bias

copy()

get_bias(j2000)

Return the current bias after applying a random-walk update.

Parameters:

j2000 (float) – Current time in Julian centuries since J2000.

Returns:

The updated bias value \(b_k\).

Return type:

float

process_covariance(dt=1.0, *, form='full')

Return random-walk covariance \(Q_b=\mathrm{diag}(\sigma_b^2\Delta t)\).

Parameters:

• dt (float)

• form (str)

Return type:

Covariance

process_psd(*, form='full')

Return continuous random-walk PSD \(\operatorname{diag}(\sigma_b^2)\).

std_bias is a random-walk standard-deviation rate, not a per-step covariance. Discretizers such as Van Loan therefore use this PSD directly and own the timestep conversion.

Parameters:

form (str)

Return type:

Covariance