Module: sunlineSEKF

Executive Summary

This module implements and tests a Switch Extended Kalman Filter in order to estimate the sunline direction.

More information can be found in the PDF Description

Message Connection Descriptions

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.

Module I/O Messages
Msg Variable Name	Msg Type	Description
navStateOutMsg	NavAttMsgPayload	name of the navigation output message containing the estimated states
filtDataOutMsg	SunlineFilterMsgPayload	name of the output filter data message
cssDataInMsg	CSSArraySensorMsgPayload	name of the CSS sensor input message
cssConfigInMsg	CSSConfigMsgPayload	name of the CSS configuration input message

Functions

void SelfInit_sunlineSEKF(sunlineSEKFConfig *configData, int64_t moduleID)

This method initializes the configData for theCSS WLS estimator. It checks to ensure that the inputs are sane and then creates the output message

Parameters:

configData – The configuration data associated with the CSS WLS estimator
moduleID – The module identifier

Returns:

void

void Reset_sunlineSEKF(sunlineSEKFConfig *configData, uint64_t callTime, int64_t moduleID)

This method resets the sunline attitude filter to an initial state and initializes the internal estimation matrices.

Parameters:

configData – The configuration data associated with the CSS estimator
callTime – The clock time at which the function was called (nanoseconds)
moduleID – The module identifier

Returns:

void

void Update_sunlineSEKF(sunlineSEKFConfig *configData, uint64_t callTime, int64_t moduleID)

This method takes the parsed CSS sensor data and outputs an estimate of the sun vector in the ADCS body frame

Parameters:

configData – The configuration data associated with the CSS estimator
callTime – The clock time at which the function was called (nanoseconds)
moduleID – The module identifier

Returns:

void

void sunlineTimeUpdate(sunlineSEKFConfig *configData, double updateTime)

This method performs the time update for the sunline kalman filter. It calls for the updated Dynamics Matrix, as well as the new states and STM. It then updates the covariance, with process noise.

Parameters:

configData – The configuration data associated with the CSS estimator
updateTime – The time that we need to fix the filter to (seconds)

Returns:

void

void sunlineMeasUpdate(sunlineSEKFConfig *configData, double updateTime)

This method performs the measurement update for the sunline kalman filter. It applies the observations in the obs vectors to the current state estimate and updates the state/covariance with that information.

Parameters:

configData – The configuration data associated with the CSS estimator
updateTime – The time that we need to fix the filter to (seconds)

Returns:

void

void sunlineStateSTMProp(double dynMat[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH], double bVec[SKF_N_STATES], double dt, double *stateInOut, double *stateTransition)

This method propagates a sunline state vector forward in time. Note that the calling parameter is updated in place to save on data copies. This also updates the STM using the dynamics matrix.

Parameters:

stateInOut –
dynMat –
bVec –
dt – time step
stateTransition –

Returns:

void

void sunlineHMatrixYMeas(double states[EKF_N_STATES_SWITCH], size_t numCSS, double cssSensorCos[MAX_N_CSS_MEAS], double sensorUseThresh, double cssNHat_B[MAX_NUM_CSS_SENSORS * 3], double *obs, double *yMeas, int *numObs, double *measMat)

This method computes the H matrix, defined by dGdX. As well as computing the innovation, difference between the measurements and the expected measurements. This methods modifies the numObs, measMat, and yMeas.

Parameters:

states –
numCSS – The total number of CSS
cssSensorCos – The list of the measurements from the CSSs
sensorUseThresh – Thresh The Threshold below which the measuremnts are not read
cssNHat_B – The normals vector for each of the CSSs
obs – Pointer to the observations
yMeas – Pointer to the innovation
numObs – Pointer to the number of observations
measMat – Point to the H measurement matrix

Returns:

void

void sunlineKalmanGain(double covarBar[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH], double hObs[MAX_N_CSS_MEAS * EKF_N_STATES_SWITCH], double qObsVal, size_t numObs, double *kalmanGain)

This method computes the Kalman gain given the measurements.

Parameters:

covarBar – The time updated covariance
hObs – The H matrix filled with the observations
qObsVal – The observation noise
numObs – The number of observations
kalmanGain – Pointer to the Kalman Gain

Returns:

void

void sunlineDynMatrix(double states[EKF_N_STATES_SWITCH], double bVec[SKF_N_STATES], double dt, double *dynMat)

This method computes the dynamics matrix, which is the derivative of the dynamics F by the state X, evaluated at the reference state. It takes in the configure data and updates this A matrix pointer called dynMat

Parameters:

states – Updated states
bVec – b vector
dt – Time step
dynMat – Pointer to the Dynamic Matrix

Returns:

void

void sunlineCKFUpdate(double xBar[EKF_N_STATES_SWITCH], double kalmanGain[EKF_N_STATES_SWITCH * MAX_N_CSS_MEAS], double covarBar[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH], double qObsVal, size_t numObs, double yObs[MAX_N_CSS_MEAS], double hObs[MAX_N_CSS_MEAS * EKF_N_STATES_SWITCH], double *x, double *covar)

This method computes the updated with a Classical Kalman Filter

Parameters:

xBar – The state after a time update
kalmanGain – The computed Kalman Gain
covarBar – The time updated covariance
qObsVal – The observation noise
numObs – The amount of CSSs that get measurements
yObs – The y vector after receiving the measurements
hObs – The H matrix filled with the observations
x – Pointer to the state error for modification
covar – Pointer to the covariance after update

Returns:

void

void sunlineSEKFUpdate(double kalmanGain[EKF_N_STATES_SWITCH * MAX_N_CSS_MEAS], double covarBar[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH], double qObsVal, size_t numObs, double yObs[MAX_N_CSS_MEAS], double hObs[MAX_N_CSS_MEAS * EKF_N_STATES_SWITCH], double *states, double *x, double *covar)

This method computes the updated with a Extended Kalman Filter

Parameters:

kalmanGain – The computed Kalman Gain
covarBar – The time updated covariance
qObsVal – The observation noise
numObs – The amount of CSSs that get measurements
yObs – The y vector after receiving the measurements
hObs – The H matrix filled with the observations
states – Pointer to the states
x – Pointer to the state error for modification
covar – Pointer to the covariance after update

Returns:

void

void sunlineSEKFSwitch(double *bVec_B, double *states, double *covar)

This method computes the dcms necessary for the switch between the two frames. It the switches the states and the covariance, and sets s2 to be the new, different vector of the body frame.

Parameters:

bVec_B – Pointer to b vector
states – Pointer to the states
covar – Pointer to the covariance

Returns:

void

void sunlineSEKFComputeDCM_BS(double sunheading[SKF_N_STATES_HALF], double bVec[SKF_N_STATES_HALF], double *dcm)

compute the DCM

Parameters:

sunheading – array of sun heading measurement states
bVec – array of bVec values
dcm – DCM pointer to be returned

Returns:

void

struct sunlineSEKFConfig

#include <sunlineSEKF.h>

Top level structure for the CSS-based Switch Extended Kalman Filter. Used to estimate the sun state in the vehicle body frame.

Public Members

NavAttMsg_C navStateOutMsg: The name of the output message

SunlineFilterMsg_C filtDataOutMsg: The name of the output filter data message

CSSArraySensorMsg_C cssDataInMsg: The name of the Input message

CSSConfigMsg_C cssConfigInMsg: [-] The name of the CSS configuration message

double qObsVal: [-] CSS instrument noise parameter

double qProcVal: [-] Process noise parameter

double dt: [s] seconds since last data epoch

double timeTag: [s] Time tag for statecovar/etc

double bVec_B[SKF_N_STATES_HALF]: [-] current vector of the b frame used to make Switch frame

double switchTresh: [-] Cosine of angle between singularity and S-frame. If close to 1, the threshold for switching frames is lower. If closer to 0.5 singularity is more largely avoided but switching is more frequent

double state[EKF_N_STATES_SWITCH]: [-] State estimate for time TimeTag

double x[EKF_N_STATES_SWITCH]: State errors

double xBar[EKF_N_STATES_SWITCH]: [-] Current mean state estimate

double covarBar[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH]: [-] Time updated covariance

double covar[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH]: [-] covariance

double stateTransition[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH]: [-] State transition Matrix

double kalmanGain[EKF_N_STATES_SWITCH * MAX_N_CSS_MEAS]: Kalman Gain

double dynMat[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH]: [-] Dynamics Matrix, A

double measMat[MAX_N_CSS_MEAS * EKF_N_STATES_SWITCH]: [-] Measurement Matrix, H

double W_BS[EKF_N_STATES_SWITCH * EKF_N_STATES_SWITCH]: [-] Switch Matrix to bring states and covariance to new S-frame when switch occurs

double obs[MAX_N_CSS_MEAS]: [-] Observation vector for frame

double yMeas[MAX_N_CSS_MEAS]: [-] Linearized measurement model data

double postFits[MAX_N_CSS_MEAS]: [-] PostFit residuals

double procNoise[(EKF_N_STATES_SWITCH - 3) * (EKF_N_STATES_SWITCH - 3)]: [-] process noise matrix

double measNoise[MAX_N_CSS_MEAS * MAX_N_CSS_MEAS]: [-] Maximally sized obs noise matrix

double cssNHat_B[MAX_NUM_CSS_SENSORS * 3]: [-] CSS normal vectors converted over to body

uint32_t numStates: [-] Number of states for this filter

size_t numObs: [-] Number of measurements this cycle

uint32_t numActiveCss: — Number of currently active CSS sensors

uint32_t numCSSTotal: [-] Count on the number of CSS we have on the spacecraft

double sensorUseThresh: — Threshold below which we discount sensors

double eKFSwitch: — Max covariance element after which the filter switches to an EKF

NavAttMsgPayload outputSunline: — Output sunline estimate data

CSSArraySensorMsgPayload cssSensorInBuffer: [-] CSS sensor data read in from message bus

BSKLogger *bskLogger: BSK Logging.