ekfInterface

Executive Summary

This class provides an interface for Extended (and Classical) Kalman Filters. Modules which inherit from it will benefit from having the core functionality of these basic filters. This class inherits from the Kalman Filter interface.

The math is derived primarily using the equations in Tapley, Schutz, and Born’s Statistical Orbit Determination textbook Chapter 4. This contains further information on this module’s mathematical framework, and the flow diagrams describing the algorithm implementations.

Virtual and Private method descriptors

The following table lists all the module input and output messages. The module msg connection is set by the user from python. The msg type contains a link to the message structure definition, while the description provides information on what this message is used for.

Interface methods which remain private

Method Name	Method Function	Protected or Private	Virtual or not
timeUpdate	filter time update	private	non-virtual
measurementUpdate	filter measurement update in the presence of measurements	private	non-virtual
computeResiduals	compute filter post fit residuals	private	non-virtual
computeKalmanGain	compute the Kalman	private	non-virtual
updateCovariance	measurement update for the covariance	private	non-virtual
ckfUpdate	update using the classical kalman filter	private	non-virtual
ekfUpdate	update using the extended kalman filter	private	non-virtual

Module assumptions and limitations

The module inherits all assumptions made while implementing a Kalman filter:

• Observability considerations

• Gaussian covariances

• Linearity limits

• and more

Otherwise, the limitations are governed by the measurement model and dynamics models relative to the required performance.

User Guide

When instantiating this filter, the user must use the FilterType enum class to chose whether this is a Linear (Classical) Kalman Filter or an Extended Kalman Filter. The Extended Kalman Filter updates the reference state at each measurement update.

This section also lists all the setters and getters that are defined by the interface.

Interface methods which remain private

Method Name	Method Function	Necessity
set/getMinimumCovarianceNormForEkf	if using an EKF, set the covariance infinite norm that will switch from CKF updates to EKFs	optional
set/getConstantRateStates	if using constant rates (not estimated by the filter) that integrate the state linearly, set this rate component	necessary

class EkfInterface : public KalmanFilter

Extended or Classical/Linear Kalman Filter base class.

Subclassed by LinearODeKF

Public Functions

void reset(uint64_t currentSimNanos) override

Reset all of the filter states, including the custom reset

Parameters:

currentSimNanos – The clock time at which the function was called (nanoseconds)

Returns:

void

void setFilterDynamicsMatrix(const std::function<const Eigen::MatrixXd(const double, const FilterStateVector&)> &dynamicsMatrixCalculator)

Get the filter dynamics matrix (A = df/dX evaluated at the reference)

Returns:

Eigen::VectorXd stateInitial

void setMinimumCovarianceNormForEkf(double infiniteNorm)

Set a minimum value (infinite norm, meaning maximal term) of the covariance before switching to Extended KF updates. This prevents divergence if the initial covariance is high and the state changes too abruptly

Parameters:

double – infiniteNorm

Returns:

void

double getMinimumCovarianceNormForEkf() const

Get the minimum value of the covariance before switching to Extended KF updates.

Returns:

double infiniteNorm