/**
 ******************************************************************************
 * @addtogroup OpenPilotModules OpenPilot Modules
 * @{
 * @addtogroup Attitude Copter Control Attitude Estimation
 * @brief Acquires sensor data and computes attitude estimate
 * Specifically updates the the @ref AttitudeState "AttitudeState" and @ref AttitudeRaw "AttitudeRaw" settings objects
 * @{
 *
 * @file       attitude.c
 * @author     The OpenPilot Team, http://www.openpilot.org Copyright (C) 2010.
 * @brief      Module to handle all comms to the AHRS on a periodic basis.
 *
 * @see        The GNU Public License (GPL) Version 3
 *
 ******************************************************************************/
/*
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 3 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
 * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
 * for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write to the Free Software Foundation, Inc.,
 * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 */

/**
 * Input objects: None, takes sensor data via pios
 * Output objects: @ref AttitudeRaw @ref AttitudeState
 *
 * This module computes an attitude estimate from the sensor data
 *
 * The module executes in its own thread.
 *
 * UAVObjects are automatically generated by the UAVObjectGenerator from
 * the object definition XML file.
 *
 * Modules have no API, all communication to other modules is done through UAVObjects.
 * However modules may use the API exposed by shared libraries.
 * See the OpenPilot wiki for more details.
 * http://www.openpilot.org/OpenPilot_Application_Architecture
 *
 */

#include <openpilot.h>

#include "attitude.h"
#include "accelsensor.h"
#include "accelstate.h"
#include "airspeedsensor.h"
#include "airspeedstate.h"
#include "attitudestate.h"
#include "attitudesettings.h"
#include "barosensor.h"
#include "flightstatus.h"
#include "gpspositionsensor.h"
#include "gpsvelocitysensor.h"
#include "gyrostate.h"
#include "gyrosensor.h"
#include "homelocation.h"
#include "magsensor.h"
#include "magstate.h"
#include "positionstate.h"
#include "ekfconfiguration.h"
#include "ekfstatevariance.h"
#include "revocalibration.h"
#include "revosettings.h"
#include "velocitystate.h"
#include "taskinfo.h"

#include "CoordinateConversions.h"

// Private constants
#define STACK_SIZE_BYTES          2048
#define TASK_PRIORITY             (tskIDLE_PRIORITY + 3)
#define FAILSAFE_TIMEOUT_MS       10

#define CALIBRATION_DELAY         4000
#define CALIBRATION_DURATION      6000
// low pass filter configuration to calculate offset
// of barometric altitude sensor
// reasoning: updates at: 10 Hz, tau= 300 s settle time
// exp(-(1/f) / tau ) ~=~ 0.9997
#define BARO_OFFSET_LOWPASS_ALPHA 0.9997f

// simple IAS to TAS aproximation - 2% increase per 1000ft
// since we do not have flowing air temperature information
#define IAS2TAS(alt) (1.0f + (0.02f * (alt) / 304.8f))

// Private types

// Private variables
static xTaskHandle attitudeTaskHandle;

static xQueueHandle gyroQueue;
static xQueueHandle accelQueue;
static xQueueHandle magQueue;
static xQueueHandle airspeedQueue;
static xQueueHandle baroQueue;
static xQueueHandle gpsQueue;
static xQueueHandle gpsVelQueue;

static AttitudeSettingsData attitudeSettings;
static HomeLocationData homeLocation;
static RevoCalibrationData revoCalibration;
static EKFConfigurationData ekfConfiguration;
static RevoSettingsData revoSettings;
static FlightStatusData flightStatus;
const uint32_t SENSOR_QUEUE_SIZE    = 10;

static bool volatile variance_error = true;
static bool volatile initialization_required = true;
static uint32_t volatile running_algorithm   = 0xffffffff; // we start with no algorithm running
static float rollPitchBiasRate = 0;

// Accel filtering
static float accel_alpha = 0;
static bool accel_filter_enabled = false;
static float accels_filtered[3];
static float grot_filtered[3];

// Private functions
static void AttitudeTask(void *parameters);

static int32_t updateAttitudeComplementary(bool first_run);
static int32_t updateAttitudeINSGPS(bool first_run, bool outdoor_mode);
static void settingsUpdatedCb(UAVObjEvent *objEv);

static int32_t getNED(GPSPositionSensorData *gpsPosition, float *NED);

static void magOffsetEstimation(MagSensorData *mag);

// check for invalid values
static inline bool invalid(float data)
{
    if (isnan(data) || isinf(data)) {
        return true;
    }
    return false;
}

// check for invalid variance values
static inline bool invalid_var(float data)
{
    if (invalid(data)) {
        return true;
    }
    if (data < 1e-15f) { // var should not be close to zero. And not negative either.
        return true;
    }
    return false;
}

/**
 * API for sensor fusion algorithms:
 * Configure(xQueueHandle gyro, xQueueHandle accel, xQueueHandle mag, xQueueHandle baro)
 *   Stores all the queues the algorithm will pull data from
 * FinalizeSensors() -- before saving the sensors modifies them based on internal state (gyro bias)
 * Update() -- queries queues and updates the attitude estiamte
 */


/**
 * Initialise the module.  Called before the start function
 * \returns 0 on success or -1 if initialisation failed
 */
int32_t AttitudeInitialize(void)
{
    GyroSensorInitialize();
    GyroStateInitialize();
    AccelSensorInitialize();
    AccelStateInitialize();
    MagSensorInitialize();
    MagStateInitialize();
    AirspeedSensorInitialize();
    AirspeedStateInitialize();
    BaroSensorInitialize();
    GPSPositionSensorInitialize();
    GPSVelocitySensorInitialize();
    AttitudeSettingsInitialize();
    AttitudeStateInitialize();
    PositionStateInitialize();
    VelocityStateInitialize();
    RevoSettingsInitialize();
    RevoCalibrationInitialize();
    EKFConfigurationInitialize();
    EKFStateVarianceInitialize();
    FlightStatusInitialize();

    // Initialize this here while we aren't setting the homelocation in GPS
    HomeLocationInitialize();

    // Initialize quaternion
    AttitudeStateData attitude;
    AttitudeStateGet(&attitude);
    attitude.q1 = 1.0f;
    attitude.q2 = 0.0f;
    attitude.q3 = 0.0f;
    attitude.q4 = 0.0f;
    AttitudeStateSet(&attitude);

    AttitudeSettingsConnectCallback(&settingsUpdatedCb);
    RevoSettingsConnectCallback(&settingsUpdatedCb);
    RevoCalibrationConnectCallback(&settingsUpdatedCb);
    HomeLocationConnectCallback(&settingsUpdatedCb);
    EKFConfigurationConnectCallback(&settingsUpdatedCb);
    FlightStatusConnectCallback(&settingsUpdatedCb);

    return 0;
}

/**
 * Start the task.  Expects all objects to be initialized by this point.
 * \returns 0 on success or -1 if initialisation failed
 */
int32_t AttitudeStart(void)
{
    // Create the queues for the sensors
    gyroQueue     = xQueueCreate(1, sizeof(UAVObjEvent));
    accelQueue    = xQueueCreate(1, sizeof(UAVObjEvent));
    magQueue      = xQueueCreate(1, sizeof(UAVObjEvent));
    airspeedQueue = xQueueCreate(1, sizeof(UAVObjEvent));
    baroQueue     = xQueueCreate(1, sizeof(UAVObjEvent));
    gpsQueue      = xQueueCreate(1, sizeof(UAVObjEvent));
    gpsVelQueue   = xQueueCreate(1, sizeof(UAVObjEvent));

    // Start main task
    xTaskCreate(AttitudeTask, (signed char *)"Attitude", STACK_SIZE_BYTES / 4, NULL, TASK_PRIORITY, &attitudeTaskHandle);
    PIOS_TASK_MONITOR_RegisterTask(TASKINFO_RUNNING_ATTITUDE, attitudeTaskHandle);
#ifdef PIOS_INCLUDE_WDG
    PIOS_WDG_RegisterFlag(PIOS_WDG_ATTITUDE);
#endif

    GyroSensorConnectQueue(gyroQueue);
    AccelSensorConnectQueue(accelQueue);
    MagSensorConnectQueue(magQueue);
    AirspeedSensorConnectQueue(airspeedQueue);
    BaroSensorConnectQueue(baroQueue);
    GPSPositionSensorConnectQueue(gpsQueue);
    GPSVelocitySensorConnectQueue(gpsVelQueue);

    return 0;
}

MODULE_INITCALL(AttitudeInitialize, AttitudeStart);

/**
 * Module thread, should not return.
 */
static void AttitudeTask(__attribute__((unused)) void *parameters)
{
    AlarmsClear(SYSTEMALARMS_ALARM_ATTITUDE);

    // Force settings update to make sure rotation loaded
    settingsUpdatedCb(NULL);

    // Wait for all the sensors be to read
    vTaskDelay(100);

    // Main task loop - TODO: make it run as delayed callback
    while (1) {
        int32_t ret_val = -1;

        bool first_run  = false;
        if (initialization_required) {
            initialization_required = false;
            first_run = true;
        }

        // This  function blocks on data queue
        switch (running_algorithm) {
        case REVOSETTINGS_FUSIONALGORITHM_COMPLEMENTARY:
            ret_val = updateAttitudeComplementary(first_run);
            break;
        case REVOSETTINGS_FUSIONALGORITHM_INS13OUTDOOR:
            ret_val = updateAttitudeINSGPS(first_run, true);
            break;
        case REVOSETTINGS_FUSIONALGORITHM_INS13INDOOR:
            ret_val = updateAttitudeINSGPS(first_run, false);
            break;
        default:
            AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_CRITICAL);
            break;
        }

        if (ret_val != 0) {
            initialization_required = true;
        }

#ifdef PIOS_INCLUDE_WDG
        PIOS_WDG_UpdateFlag(PIOS_WDG_ATTITUDE);
#endif
    }
}

static inline void apply_accel_filter(const float *raw, float *filtered)
{
    if (accel_filter_enabled) {
        filtered[0] = filtered[0] * accel_alpha + raw[0] * (1 - accel_alpha);
        filtered[1] = filtered[1] * accel_alpha + raw[1] * (1 - accel_alpha);
        filtered[2] = filtered[2] * accel_alpha + raw[2] * (1 - accel_alpha);
    } else {
        filtered[0] = raw[0];
        filtered[1] = raw[1];
        filtered[2] = raw[2];
    }
}

float accel_mag;
float qmag;
float attitudeDt;
float mag_err[3];

static int32_t updateAttitudeComplementary(bool first_run)
{
    UAVObjEvent ev;
    GyroSensorData gyroSensorData;
    GyroStateData gyroStateData;
    AccelSensorData accelSensorData;
    static int32_t timeval;
    float dT;
    static uint8_t init = 0;
    static float gyro_bias[3] = { 0, 0, 0 };
    static bool magCalibrated = true;
    static uint32_t initStartupTime = 0;

    // Wait until the AttitudeRaw object is updated, if a timeout then go to failsafe
    if (xQueueReceive(gyroQueue, &ev, FAILSAFE_TIMEOUT_MS / portTICK_RATE_MS) != pdTRUE ||
        xQueueReceive(accelQueue, &ev, 1 / portTICK_RATE_MS) != pdTRUE) {
        // When one of these is updated so should the other
        // Do not set attitude timeout warnings in simulation mode
        if (!AttitudeStateReadOnly()) {
            AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_WARNING);
            return -1;
        }
    }

    AccelSensorGet(&accelSensorData);

// TODO: put in separate filter
    AccelStateData accelState;
    accelState.x = accelSensorData.x;
    accelState.y = accelSensorData.y;
    accelState.z = accelSensorData.z;
    AccelStateSet(&accelState);

    // During initialization and
    if (first_run) {
#if defined(PIOS_INCLUDE_HMC5883)
        // To initialize we need a valid mag reading
        if (xQueueReceive(magQueue, &ev, 0 / portTICK_RATE_MS) != pdTRUE) {
            return -1;
        }
        MagSensorData magData;
        MagSensorGet(&magData);
#else
        MagSensorData magData;
        magData.x = 100.0f;
        magData.y = 0.0f;
        magData.z = 0.0f;
#endif
        float magBias[3];
        RevoCalibrationmag_biasGet(magBias);
        // don't trust Mag for initial orientation if it has not been calibrated
        if (magBias[0] < 1e-6f && magBias[1] < 1e-6f && magBias[2] < 1e-6f) {
            magCalibrated = false;
            magData.x     = 100.0f;
            magData.y     = 0.0f;
            magData.z     = 0.0f;
        }
        AttitudeStateData attitudeState;
        AttitudeStateGet(&attitudeState);
        init = 0;

        // Set initial attitude. Use accels to determine roll and pitch, rotate magnetic measurement accordingly,
        // so pseudo "north" vector can be estimated even if the board is not level
        attitudeState.Roll = atan2f(-accelSensorData.y, -accelSensorData.z);
        float zn  = cosf(attitudeState.Roll) * magData.z + sinf(attitudeState.Roll) * magData.y;
        float yn  = cosf(attitudeState.Roll) * magData.y - sinf(attitudeState.Roll) * magData.z;

        // rotate accels z vector according to roll
        float azn = cosf(attitudeState.Roll) * accelSensorData.z + sinf(attitudeState.Roll) * accelSensorData.y;
        attitudeState.Pitch = atan2f(accelSensorData.x, -azn);

        float xn  = cosf(attitudeState.Pitch) * magData.x + sinf(attitudeState.Pitch) * zn;

        attitudeState.Yaw = atan2f(-yn, xn);
        // TODO: This is still a hack
        // Put this in a proper generic function in CoordinateConversion.c
        // should take 4 vectors: g (0,0,-9.81), accels, Be (or 1,0,0 if no home loc) and magnetometers (or 1,0,0 if no mags)
        // should calculate the rotation in 3d space using proper cross product math
        // SUBTODO: formulate the math required

        attitudeState.Roll  = RAD2DEG(attitudeState.Roll);
        attitudeState.Pitch = RAD2DEG(attitudeState.Pitch);
        attitudeState.Yaw   = RAD2DEG(attitudeState.Yaw);

        RPY2Quaternion(&attitudeState.Roll, &attitudeState.q1);
        AttitudeStateSet(&attitudeState);

        timeval = PIOS_DELAY_GetRaw();

        // wait calibration_delay only at powerup
        if (xTaskGetTickCount() < 3000) {
            initStartupTime = 0;
        } else {
            initStartupTime = xTaskGetTickCount() - CALIBRATION_DELAY;
        }

        // Zero gyro bias
        // This is really needed after updating calibration settings.
        gyro_bias[0] = 0.0f;
        gyro_bias[1] = 0.0f;
        gyro_bias[2] = 0.0f;
        return 0;
    }

    if ((xTaskGetTickCount() - initStartupTime < CALIBRATION_DURATION + CALIBRATION_DELAY) &&
        (xTaskGetTickCount() - initStartupTime > CALIBRATION_DELAY)) {
        // For first CALIBRATION_DURATION seconds after CALIBRATION_DELAY from startup
        // Zero gyro bias assuming it is steady, smoothing the gyro input value applying rollPitchBiasRate.
        attitudeSettings.AccelKp     = 1.0f;
        attitudeSettings.AccelKi     = 0.0f;
        attitudeSettings.YawBiasRate = 0.23f;
        accel_filter_enabled   = false;
        rollPitchBiasRate      = 0.01f;
        attitudeSettings.MagKp = magCalibrated ? 1.0f : 0.0f;
        init = 0;
    } else if ((attitudeSettings.ZeroDuringArming == ATTITUDESETTINGS_ZERODURINGARMING_TRUE) && (flightStatus.Armed == FLIGHTSTATUS_ARMED_ARMING)) {
        attitudeSettings.AccelKp     = 1.0f;
        attitudeSettings.AccelKi     = 0.0f;
        attitudeSettings.YawBiasRate = 0.23f;
        accel_filter_enabled   = false;
        rollPitchBiasRate      = 0.01f;
        attitudeSettings.MagKp = magCalibrated ? 1.0f : 0.0f;
        init = 0;
    } else if (init == 0) {
        // Reload settings (all the rates)
        AttitudeSettingsGet(&attitudeSettings);
        rollPitchBiasRate = 0.0f;
        if (accel_alpha > 0.0f) {
            accel_filter_enabled = true;
        }
        init = 1;
    }

    GyroSensorGet(&gyroSensorData);
    gyroStateData.x = gyroSensorData.x;
    gyroStateData.y = gyroSensorData.y;
    gyroStateData.z = gyroSensorData.z;

    // Compute the dT using the cpu clock
    dT = PIOS_DELAY_DiffuS(timeval) / 1000000.0f;
    timeval = PIOS_DELAY_GetRaw();

    float q[4];

    AttitudeStateData attitudeState;
    AttitudeStateGet(&attitudeState);

    float grot[3];
    float accel_err[3];

    // Get the current attitude estimate
    quat_copy(&attitudeState.q1, q);

    // Apply smoothing to accel values, to reduce vibration noise before main calculations.
    apply_accel_filter((const float *)&accelSensorData.x, accels_filtered);

    // Rotate gravity to body frame and cross with accels
    grot[0] = -(2.0f * (q[1] * q[3] - q[0] * q[2]));
    grot[1] = -(2.0f * (q[2] * q[3] + q[0] * q[1]));
    grot[2] = -(q[0] * q[0] - q[1] * q[1] - q[2] * q[2] + q[3] * q[3]);

    apply_accel_filter(grot, grot_filtered);

    CrossProduct((const float *)accels_filtered, (const float *)grot_filtered, accel_err);

    // Account for accel magnitude
    accel_mag = accels_filtered[0] * accels_filtered[0] + accels_filtered[1] * accels_filtered[1] + accels_filtered[2] * accels_filtered[2];
    accel_mag = sqrtf(accel_mag);

    float grot_mag;
    if (accel_filter_enabled) {
        grot_mag = sqrtf(grot_filtered[0] * grot_filtered[0] + grot_filtered[1] * grot_filtered[1] + grot_filtered[2] * grot_filtered[2]);
    } else {
        grot_mag = 1.0f;
    }

    // TODO! check grot_mag & accel vector magnitude values for correctness.

    accel_err[0] /= (accel_mag * grot_mag);
    accel_err[1] /= (accel_mag * grot_mag);
    accel_err[2] /= (accel_mag * grot_mag);


    if (xQueueReceive(magQueue, &ev, 0) != pdTRUE) {
        // Rotate gravity to body frame and cross with accels
        float brot[3];
        float Rbe[3][3];
        MagSensorData mag;

        Quaternion2R(q, Rbe);
        MagSensorGet(&mag);

// TODO: separate filter!
        if (revoCalibration.MagBiasNullingRate > 0) {
            magOffsetEstimation(&mag);
        }
        MagStateData mags;
        mags.x = mag.x;
        mags.y = mag.y;
        mags.z = mag.z;
        MagStateSet(&mags);

        // If the mag is producing bad data don't use it (normally bad calibration)
        if (!isnan(mag.x) && !isinf(mag.x) && !isnan(mag.y) && !isinf(mag.y) && !isnan(mag.z) && !isinf(mag.z)) {
            rot_mult(Rbe, homeLocation.Be, brot);

            float mag_len = sqrtf(mag.x * mag.x + mag.y * mag.y + mag.z * mag.z);
            mag.x   /= mag_len;
            mag.y   /= mag_len;
            mag.z   /= mag_len;

            float bmag = sqrtf(brot[0] * brot[0] + brot[1] * brot[1] + brot[2] * brot[2]);
            brot[0] /= bmag;
            brot[1] /= bmag;
            brot[2] /= bmag;

            // Only compute if neither vector is null
            if (bmag < 1.0f || mag_len < 1.0f) {
                mag_err[0] = mag_err[1] = mag_err[2] = 0.0f;
            } else {
                CrossProduct((const float *)&mag.x, (const float *)brot, mag_err);
            }
        }
    } else {
        mag_err[0] = mag_err[1] = mag_err[2] = 0.0f;
    }

    // Accumulate integral of error.  Scale here so that units are (deg/s) but Ki has units of s
    // Correct rates based on integral coefficient
    gyroStateData.x -= gyro_bias[0];
    gyroStateData.y -= gyro_bias[1];
    gyroStateData.z -= gyro_bias[2];

    gyro_bias[0]    -= accel_err[0] * attitudeSettings.AccelKi - (gyroStateData.x) * rollPitchBiasRate;
    gyro_bias[1]    -= accel_err[1] * attitudeSettings.AccelKi - (gyroStateData.y) * rollPitchBiasRate;
    gyro_bias[2]    -= -mag_err[2] * attitudeSettings.MagKi - (gyroStateData.z) * rollPitchBiasRate;

    // save gyroscope state
    GyroStateSet(&gyroStateData);

    // Correct rates based on proportional coefficient
    gyroStateData.x += accel_err[0] * attitudeSettings.AccelKp / dT;
    gyroStateData.y += accel_err[1] * attitudeSettings.AccelKp / dT;
    gyroStateData.z += accel_err[2] * attitudeSettings.AccelKp / dT + mag_err[2] * attitudeSettings.MagKp / dT;

    // Work out time derivative from INSAlgo writeup
    // Also accounts for the fact that gyros are in deg/s
    float qdot[4];
    qdot[0] = DEG2RAD(-q[1] * gyroStateData.x - q[2] * gyroStateData.y - q[3] * gyroStateData.z) * dT / 2;
    qdot[1] = DEG2RAD(q[0] * gyroStateData.x - q[3] * gyroStateData.y + q[2] * gyroStateData.z) * dT / 2;
    qdot[2] = DEG2RAD(q[3] * gyroStateData.x + q[0] * gyroStateData.y - q[1] * gyroStateData.z) * dT / 2;
    qdot[3] = DEG2RAD(-q[2] * gyroStateData.x + q[1] * gyroStateData.y + q[0] * gyroStateData.z) * dT / 2;

    // Take a time step
    q[0]    = q[0] + qdot[0];
    q[1]    = q[1] + qdot[1];
    q[2]    = q[2] + qdot[2];
    q[3]    = q[3] + qdot[3];

    if (q[0] < 0.0f) {
        q[0] = -q[0];
        q[1] = -q[1];
        q[2] = -q[2];
        q[3] = -q[3];
    }

    // Renomalize
    qmag = sqrtf(q[0] * q[0] + q[1] * q[1] + q[2] * q[2] + q[3] * q[3]);
    q[0] = q[0] / qmag;
    q[1] = q[1] / qmag;
    q[2] = q[2] / qmag;
    q[3] = q[3] / qmag;

    // If quaternion has become inappropriately short or is nan reinit.
    // THIS SHOULD NEVER ACTUALLY HAPPEN
    if ((fabsf(qmag) < 1.0e-3f) || isnan(qmag)) {
        q[0] = 1.0f;
        q[1] = 0.0f;
        q[2] = 0.0f;
        q[3] = 0.0f;
    }

    quat_copy(q, &attitudeState.q1);

    // Convert into eueler degrees (makes assumptions about RPY order)
    Quaternion2RPY(&attitudeState.q1, &attitudeState.Roll);

    AttitudeStateSet(&attitudeState);

    // Flush these queues for avoid errors
    xQueueReceive(baroQueue, &ev, 0);
    if (xQueueReceive(gpsQueue, &ev, 0) == pdTRUE && homeLocation.Set == HOMELOCATION_SET_TRUE) {
        float NED[3];
        // Transform the GPS position into NED coordinates
        GPSPositionSensorData gpsPosition;
        GPSPositionSensorGet(&gpsPosition);
        getNED(&gpsPosition, NED);

        PositionStateData positionState;
        PositionStateGet(&positionState);
        positionState.North = NED[0];
        positionState.East  = NED[1];
        positionState.Down  = NED[2];
        PositionStateSet(&positionState);
    }

    if (xQueueReceive(gpsVelQueue, &ev, 0) == pdTRUE) {
        // Transform the GPS position into NED coordinates
        GPSVelocitySensorData gpsVelocity;
        GPSVelocitySensorGet(&gpsVelocity);

        VelocityStateData velocityState;
        VelocityStateGet(&velocityState);
        velocityState.North = gpsVelocity.North;
        velocityState.East  = gpsVelocity.East;
        velocityState.Down  = gpsVelocity.Down;
        VelocityStateSet(&velocityState);
    }

    if (xQueueReceive(airspeedQueue, &ev, 0) == pdTRUE) {
        // Calculate true airspeed from indicated airspeed
        AirspeedSensorData airspeedSensor;
        AirspeedSensorGet(&airspeedSensor);

        AirspeedStateData airspeed;
        AirspeedStateGet(&airspeed);

        PositionStateData positionState;
        PositionStateGet(&positionState);

        if (airspeedSensor.SensorConnected == AIRSPEEDSENSOR_SENSORCONNECTED_TRUE) {
            // we have airspeed available
            airspeed.CalibratedAirspeed = airspeedSensor.CalibratedAirspeed;
            airspeed.TrueAirspeed = airspeed.CalibratedAirspeed * IAS2TAS(homeLocation.Altitude - positionState.Down);
            AirspeedStateSet(&airspeed);
        }
    }

    if (!init && flightStatus.Armed == FLIGHTSTATUS_ARMED_DISARMED) {
        AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_ERROR);
    } else if (variance_error) {
        AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_CRITICAL);
    } else {
        AlarmsClear(SYSTEMALARMS_ALARM_ATTITUDE);
    }


    return 0;
}

#include "insgps.h"
int32_t ins_failed = 0;
extern struct NavStruct Nav;
int32_t init_stage = 0;

/**
 * @brief Use the INSGPS fusion algorithm in either indoor or outdoor mode (use GPS)
 * @params[in] first_run This is the first run so trigger reinitialization
 * @params[in] outdoor_mode If true use the GPS for position, if false weakly pull to (0,0)
 * @return 0 for success, -1 for failure
 */
static int32_t updateAttitudeINSGPS(bool first_run, bool outdoor_mode)
{
    UAVObjEvent ev;
    GyroSensorData gyroSensorData;
    AccelSensorData accelSensorData;
    MagStateData magData;
    AirspeedSensorData airspeedData;
    BaroSensorData baroData;
    GPSPositionSensorData gpsData;
    GPSVelocitySensorData gpsVelData;

    static bool mag_updated = false;
    static bool baro_updated;
    static bool airspeed_updated;
    static bool gps_updated;
    static bool gps_vel_updated;

    static bool value_error = false;

    static float baroOffset = 0.0f;

    static uint32_t ins_last_time = 0;
    static bool inited;

    float NED[3]     = { 0.0f, 0.0f, 0.0f };
    float vel[3]     = { 0.0f, 0.0f, 0.0f };
    float zeros[3]   = { 0.0f, 0.0f, 0.0f };

    // Perform the update
    uint16_t sensors = 0;
    float dT;

    // Wait until the gyro and accel object is updated, if a timeout then go to failsafe
    if ((xQueueReceive(gyroQueue, &ev, FAILSAFE_TIMEOUT_MS / portTICK_RATE_MS) != pdTRUE) ||
        (xQueueReceive(accelQueue, &ev, 1 / portTICK_RATE_MS) != pdTRUE)) {
        // Do not set attitude timeout warnings in simulation mode
        if (!AttitudeStateReadOnly()) {
            AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_WARNING);
            return -1;
        }
    }

    if (inited) {
        mag_updated      = 0;
        baro_updated     = 0;
        airspeed_updated = 0;
        gps_updated      = 0;
        gps_vel_updated  = 0;
    }

    if (first_run) {
        inited           = false;
        init_stage       = 0;

        mag_updated      = 0;
        baro_updated     = 0;
        airspeed_updated = 0;
        gps_updated      = 0;
        gps_vel_updated  = 0;

        ins_last_time    = PIOS_DELAY_GetRaw();

        return 0;
    }

    mag_updated      |= (xQueueReceive(magQueue, &ev, 0 / portTICK_RATE_MS) == pdTRUE);
    baro_updated     |= xQueueReceive(baroQueue, &ev, 0 / portTICK_RATE_MS) == pdTRUE;
    airspeed_updated |= xQueueReceive(airspeedQueue, &ev, 0 / portTICK_RATE_MS) == pdTRUE;

    // Check if we are running simulation
    if (!GPSPositionSensorReadOnly()) {
        gps_updated |= (xQueueReceive(gpsQueue, &ev, 0 / portTICK_RATE_MS) == pdTRUE) && outdoor_mode;
    } else {
        gps_updated |= pdTRUE && outdoor_mode;
    }

    if (!GPSVelocitySensorReadOnly()) {
        gps_vel_updated |= (xQueueReceive(gpsVelQueue, &ev, 0 / portTICK_RATE_MS) == pdTRUE) && outdoor_mode;
    } else {
        gps_vel_updated |= pdTRUE && outdoor_mode;
    }

    // Get most recent data
    GyroSensorGet(&gyroSensorData);
    AccelSensorGet(&accelSensorData);
// TODO: separate filter!
    if (mag_updated) {
        MagSensorData mags;
        MagSensorGet(&mags);
        if (revoCalibration.MagBiasNullingRate > 0) {
            magOffsetEstimation(&mags);
        }
        magData.x = mags.x;
        magData.y = mags.y;
        magData.z = mags.z;
        MagStateSet(&magData);
    } else {
        MagStateGet(&magData);
    }

    BaroSensorGet(&baroData);
    AirspeedSensorGet(&airspeedData);
    GPSPositionSensorGet(&gpsData);
    GPSVelocitySensorGet(&gpsVelData);

// TODO: put in separate filter
    AccelStateData accelState;
    accelState.x = accelSensorData.x;
    accelState.y = accelSensorData.y;
    accelState.z = accelSensorData.z;
    AccelStateSet(&accelState);


    value_error = false;
    // safety checks
    if (invalid(gyroSensorData.x) ||
        invalid(gyroSensorData.y) ||
        invalid(gyroSensorData.z) ||
        invalid(accelSensorData.x) ||
        invalid(accelSensorData.y) ||
        invalid(accelSensorData.z)) {
        // cannot run process update, raise error!
        AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_ERROR);
        return 0;
    }

    if (invalid(magData.x) ||
        invalid(magData.y) ||
        invalid(magData.z)) {
        // magnetometers can be ignored for a while
        mag_updated = false;
        value_error = true;
    }

    // Don't require HomeLocation.Set to be true but at least require a mag configuration (allows easily
    // switching between indoor and outdoor mode with Set = false)
    if ((homeLocation.Be[0] * homeLocation.Be[0] + homeLocation.Be[1] * homeLocation.Be[1] + homeLocation.Be[2] * homeLocation.Be[2] < 1e-5f)) {
        mag_updated = false;
        value_error = true;
    }

    if (invalid(baroData.Altitude)) {
        baro_updated = false;
        value_error  = true;
    }

    if (invalid(airspeedData.CalibratedAirspeed)) {
        airspeed_updated = false;
        value_error = true;
    }

    if (invalid(gpsData.Altitude)) {
        gps_updated = false;
        value_error = true;
    }

    if (invalid_var(ekfConfiguration.R[EKFCONFIGURATION_R_GPSPOSNORTH]) ||
        invalid_var(ekfConfiguration.R[EKFCONFIGURATION_R_GPSPOSEAST]) ||
        invalid_var(ekfConfiguration.R[EKFCONFIGURATION_R_GPSPOSDOWN]) ||
        invalid_var(ekfConfiguration.R[EKFCONFIGURATION_R_GPSVELNORTH]) ||
        invalid_var(ekfConfiguration.R[EKFCONFIGURATION_R_GPSVELEAST]) ||
        invalid_var(ekfConfiguration.R[EKFCONFIGURATION_R_GPSVELDOWN])) {
        gps_updated = false;
        value_error = true;
    }

    if (invalid(gpsVelData.North) ||
        invalid(gpsVelData.East) ||
        invalid(gpsVelData.Down)) {
        gps_vel_updated = false;
        value_error     = true;
    }

    // Discard airspeed if sensor not connected
    if (airspeedData.SensorConnected != AIRSPEEDSENSOR_SENSORCONNECTED_TRUE) {
        airspeed_updated = false;
    }

    // Have a minimum requirement for gps usage
    if ((gpsData.Satellites < 7) ||
        (gpsData.PDOP > 4.0f) ||
        (gpsData.Latitude == 0 && gpsData.Longitude == 0) ||
        (homeLocation.Set != HOMELOCATION_SET_TRUE)) {
        gps_updated     = false;
        gps_vel_updated = false;
    }

    if (!inited) {
        AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_ERROR);
    } else if (value_error) {
        AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_CRITICAL);
    } else if (variance_error) {
        AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_CRITICAL);
    } else if (outdoor_mode && gpsData.Satellites < 7) {
        AlarmsSet(SYSTEMALARMS_ALARM_ATTITUDE, SYSTEMALARMS_ALARM_ERROR);
    } else {
        AlarmsClear(SYSTEMALARMS_ALARM_ATTITUDE);
    }

    dT = PIOS_DELAY_DiffuS(ins_last_time) / 1.0e6f;
    ins_last_time = PIOS_DELAY_GetRaw();

    // This should only happen at start up or at mode switches
    if (dT > 0.01f) {
        dT = 0.01f;
    } else if (dT <= 0.001f) {
        dT = 0.001f;
    }

    if (!inited && mag_updated && baro_updated && (gps_updated || !outdoor_mode) && !variance_error) {
        // Don't initialize until all sensors are read
        if (init_stage == 0) {
            // Reset the INS algorithm
            INSGPSInit();
            INSSetMagVar((float[3]) { ekfConfiguration.R[EKFCONFIGURATION_R_MAGX],
                                      ekfConfiguration.R[EKFCONFIGURATION_R_MAGY],
                                      ekfConfiguration.R[EKFCONFIGURATION_R_MAGZ] }
                         );
            INSSetAccelVar((float[3]) { ekfConfiguration.Q[EKFCONFIGURATION_Q_ACCELX],
                                        ekfConfiguration.Q[EKFCONFIGURATION_Q_ACCELY],
                                        ekfConfiguration.Q[EKFCONFIGURATION_Q_ACCELZ] }
                           );
            INSSetGyroVar((float[3]) { ekfConfiguration.Q[EKFCONFIGURATION_Q_GYROX],
                                       ekfConfiguration.Q[EKFCONFIGURATION_Q_GYROY],
                                       ekfConfiguration.Q[EKFCONFIGURATION_Q_GYROZ] }
                          );
            INSSetGyroBiasVar((float[3]) { ekfConfiguration.Q[EKFCONFIGURATION_Q_GYRODRIFTX],
                                           ekfConfiguration.Q[EKFCONFIGURATION_Q_GYRODRIFTY],
                                           ekfConfiguration.Q[EKFCONFIGURATION_Q_GYRODRIFTZ] }
                              );
            INSSetBaroVar(ekfConfiguration.R[EKFCONFIGURATION_R_BAROZ]);

            // Initialize the gyro bias
            float gyro_bias[3] = { 0.0f, 0.0f, 0.0f };
            INSSetGyroBias(gyro_bias);

            float pos[3] = { 0.0f, 0.0f, 0.0f };

            if (outdoor_mode) {
                GPSPositionSensorData gpsPosition;
                GPSPositionSensorGet(&gpsPosition);

                // Transform the GPS position into NED coordinates
                getNED(&gpsPosition, pos);

                // Initialize barometric offset to current GPS NED coordinate
                baroOffset = -pos[2] - baroData.Altitude;
            } else {
                // Initialize barometric offset to homelocation altitude
                baroOffset = -baroData.Altitude;
                pos[2]     = -(baroData.Altitude + baroOffset);
            }

            // xQueueReceive(magQueue, &ev, 100 / portTICK_RATE_MS);
            // MagSensorGet(&magData);

            AttitudeStateData attitudeState;
            AttitudeStateGet(&attitudeState);

            // Set initial attitude. Use accels to determine roll and pitch, rotate magnetic measurement accordingly,
            // so pseudo "north" vector can be estimated even if the board is not level
            attitudeState.Roll = atan2f(-accelSensorData.y, -accelSensorData.z);
            float zn  = cosf(attitudeState.Roll) * magData.z + sinf(attitudeState.Roll) * magData.y;
            float yn  = cosf(attitudeState.Roll) * magData.y - sinf(attitudeState.Roll) * magData.z;

            // rotate accels z vector according to roll
            float azn = cosf(attitudeState.Roll) * accelSensorData.z + sinf(attitudeState.Roll) * accelSensorData.y;
            attitudeState.Pitch = atan2f(accelSensorData.x, -azn);

            float xn  = cosf(attitudeState.Pitch) * magData.x + sinf(attitudeState.Pitch) * zn;

            attitudeState.Yaw = atan2f(-yn, xn);
            // TODO: This is still a hack
            // Put this in a proper generic function in CoordinateConversion.c
            // should take 4 vectors: g (0,0,-9.81), accels, Be (or 1,0,0 if no home loc) and magnetometers (or 1,0,0 if no mags)
            // should calculate the rotation in 3d space using proper cross product math
            // SUBTODO: formulate the math required

            attitudeState.Roll  = RAD2DEG(attitudeState.Roll);
            attitudeState.Pitch = RAD2DEG(attitudeState.Pitch);
            attitudeState.Yaw   = RAD2DEG(attitudeState.Yaw);

            RPY2Quaternion(&attitudeState.Roll, &attitudeState.q1);
            AttitudeStateSet(&attitudeState);

            float q[4] = { attitudeState.q1, attitudeState.q2, attitudeState.q3, attitudeState.q4 };
            INSSetState(pos, zeros, q, zeros, zeros);

            INSResetP(ekfConfiguration.P);
        } else {
            // Run prediction a bit before any corrections

            // Because the sensor module remove the bias we need to add it
            // back in here so that the INS algorithm can track it correctly
            float gyros[3] = { DEG2RAD(gyroSensorData.x), DEG2RAD(gyroSensorData.y), DEG2RAD(gyroSensorData.z) };
            INSStatePrediction(gyros, &accelSensorData.x, dT);

            AttitudeStateData attitude;
            AttitudeStateGet(&attitude);
            attitude.q1 = Nav.q[0];
            attitude.q2 = Nav.q[1];
            attitude.q3 = Nav.q[2];
            attitude.q4 = Nav.q[3];
            Quaternion2RPY(&attitude.q1, &attitude.Roll);
            AttitudeStateSet(&attitude);
        }

        init_stage++;
        if (init_stage > 10) {
            inited = true;
        }

        return 0;
    }

    if (!inited) {
        return 0;
    }

    // Because the sensor module remove the bias we need to add it
    // back in here so that the INS algorithm can track it correctly
    float gyros[3] = { DEG2RAD(gyroSensorData.x), DEG2RAD(gyroSensorData.y), DEG2RAD(gyroSensorData.z) };

    // Advance the state estimate
    INSStatePrediction(gyros, &accelSensorData.x, dT);

    // Copy the attitude into the UAVO
    AttitudeStateData attitude;
    AttitudeStateGet(&attitude);
    attitude.q1 = Nav.q[0];
    attitude.q2 = Nav.q[1];
    attitude.q3 = Nav.q[2];
    attitude.q4 = Nav.q[3];
    Quaternion2RPY(&attitude.q1, &attitude.Roll);
    AttitudeStateSet(&attitude);

    // Advance the covariance estimate
    INSCovariancePrediction(dT);

    if (mag_updated) {
        sensors |= MAG_SENSORS;
    }

    if (baro_updated) {
        sensors |= BARO_SENSOR;
    }

    INSSetMagNorth(homeLocation.Be);

    if (gps_updated && outdoor_mode) {
        INSSetPosVelVar((float[3]) { ekfConfiguration.R[EKFCONFIGURATION_R_GPSPOSNORTH],
                                     ekfConfiguration.R[EKFCONFIGURATION_R_GPSPOSEAST],
                                     ekfConfiguration.R[EKFCONFIGURATION_R_GPSPOSDOWN] },
                        (float[3]) { ekfConfiguration.R[EKFCONFIGURATION_R_GPSVELNORTH],
                                     ekfConfiguration.R[EKFCONFIGURATION_R_GPSVELEAST],
                                     ekfConfiguration.R[EKFCONFIGURATION_R_GPSVELDOWN] }
                        );
        sensors |= POS_SENSORS;

        if (0) { // Old code to take horizontal velocity from GPS Position update
            sensors |= HORIZ_SENSORS;
            vel[0]   = gpsData.Groundspeed * cosf(DEG2RAD(gpsData.Heading));
            vel[1]   = gpsData.Groundspeed * sinf(DEG2RAD(gpsData.Heading));
            vel[2]   = 0.0f;
        }
        // Transform the GPS position into NED coordinates
        getNED(&gpsData, NED);

        // Track barometric altitude offset with a low pass filter
        baroOffset = BARO_OFFSET_LOWPASS_ALPHA * baroOffset +
                     (1.0f - BARO_OFFSET_LOWPASS_ALPHA)
                     * (-NED[2] - baroData.Altitude);
    } else if (!outdoor_mode) {
        INSSetPosVelVar((float[3]) { ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSPOSINDOOR],
                                     ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSPOSINDOOR],
                                     ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSPOSINDOOR] },
                        (float[3]) { ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSVELINDOOR],
                                     ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSVELINDOOR],
                                     ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSVELINDOOR] }
                        );
        vel[0]   = vel[1] = vel[2] = 0.0f;
        NED[0]   = NED[1] = 0.0f;
        NED[2]   = -(baroData.Altitude + baroOffset);
        sensors |= HORIZ_SENSORS | HORIZ_POS_SENSORS;
        sensors |= POS_SENSORS | VERT_SENSORS;
    }

    if (gps_vel_updated && outdoor_mode) {
        sensors |= HORIZ_SENSORS | VERT_SENSORS;
        vel[0]   = gpsVelData.North;
        vel[1]   = gpsVelData.East;
        vel[2]   = gpsVelData.Down;
    }

    if (airspeed_updated) {
        // we have airspeed available
        AirspeedStateData airspeed;
        AirspeedStateGet(&airspeed);

        airspeed.CalibratedAirspeed = airspeedData.CalibratedAirspeed;
        airspeed.TrueAirspeed = airspeed.CalibratedAirspeed * IAS2TAS(homeLocation.Altitude - Nav.Pos[2]);
        AirspeedStateSet(&airspeed);

        if (!gps_vel_updated && !gps_updated) {
            // feed airspeed into EKF, treat wind as 1e2 variance
            sensors |= HORIZ_SENSORS | VERT_SENSORS;
            INSSetPosVelVar((float[3]) { ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSPOSINDOOR],
                                         ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSPOSINDOOR],
                                         ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSPOSINDOOR] },
                            (float[3]) { ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSVELAIRSPEED],
                                         ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSVELAIRSPEED],
                                         ekfConfiguration.FakeR[EKFCONFIGURATION_FAKER_FAKEGPSVELAIRSPEED] }
                            );
            // rotate airspeed vector into NED frame - airspeed is measured in X axis only
            float R[3][3];
            Quaternion2R(Nav.q, R);
            float vtas[3] = { airspeed.TrueAirspeed, 0.0f, 0.0f };
            rot_mult(R, vtas, vel);
        }
    }

    /*
     * TODO: Need to add a general sanity check for all the inputs to make sure their kosher
     * although probably should occur within INS itself
     */
    if (sensors) {
        INSCorrection(&magData.x, NED, vel, (baroData.Altitude + baroOffset), sensors);
    }

    // Copy the position and velocity into the UAVO
    PositionStateData positionState;
    PositionStateGet(&positionState);
    positionState.North = Nav.Pos[0];
    positionState.East  = Nav.Pos[1];
    positionState.Down  = Nav.Pos[2];
    PositionStateSet(&positionState);

    VelocityStateData velocityState;
    VelocityStateGet(&velocityState);
    velocityState.North = Nav.Vel[0];
    velocityState.East  = Nav.Vel[1];
    velocityState.Down  = Nav.Vel[2];
    VelocityStateSet(&velocityState);

    GyroStateData gyroState;
    gyroState.x = RAD2DEG(gyros[0] - RAD2DEG(Nav.gyro_bias[0]));
    gyroState.y = RAD2DEG(gyros[1] - RAD2DEG(Nav.gyro_bias[1]));
    gyroState.z = RAD2DEG(gyros[2] - RAD2DEG(Nav.gyro_bias[2]));
    GyroStateSet(&gyroState);

    EKFStateVarianceData vardata;
    EKFStateVarianceGet(&vardata);
    INSGetP(vardata.P);
    EKFStateVarianceSet(&vardata);

    return 0;
}

/**
 * @brief Convert the GPS LLA position into NED coordinates
 * @note this method uses a taylor expansion around the home coordinates
 * to convert to NED which allows it to be done with all floating
 * calculations
 * @param[in] Current GPS coordinates
 * @param[out] NED frame coordinates
 * @returns 0 for success, -1 for failure
 */
float T[3];
static int32_t getNED(GPSPositionSensorData *gpsPosition, float *NED)
{
    float dL[3] = { DEG2RAD((gpsPosition->Latitude - homeLocation.Latitude) / 10.0e6f),
                    DEG2RAD((gpsPosition->Longitude - homeLocation.Longitude) / 10.0e6f),
                    (gpsPosition->Altitude + gpsPosition->GeoidSeparation - homeLocation.Altitude) };

    NED[0] = T[0] * dL[0];
    NED[1] = T[1] * dL[1];
    NED[2] = T[2] * dL[2];

    return 0;
}

static void settingsUpdatedCb(UAVObjEvent *ev)
{
    if (ev == NULL || ev->obj == FlightStatusHandle()) {
        FlightStatusGet(&flightStatus);
    }
    if (ev == NULL || ev->obj == RevoCalibrationHandle()) {
        RevoCalibrationGet(&revoCalibration);
    }
    // change of these settings require reinitialization of the EKF
    // when an error flag has been risen, we also listen to flightStatus updates,
    // since we are waiting for the system to get disarmed so we can reinitialize safely.
    if (ev == NULL ||
        ev->obj == EKFConfigurationHandle() ||
        ev->obj == RevoSettingsHandle() ||
        (variance_error == true && ev->obj == FlightStatusHandle())
        ) {
        bool error = false;

        EKFConfigurationGet(&ekfConfiguration);
        int t;
        for (t = 0; t < EKFCONFIGURATION_P_NUMELEM; t++) {
            if (invalid_var(ekfConfiguration.P[t])) {
                error = true;
            }
        }
        for (t = 0; t < EKFCONFIGURATION_Q_NUMELEM; t++) {
            if (invalid_var(ekfConfiguration.Q[t])) {
                error = true;
            }
        }
        for (t = 0; t < EKFCONFIGURATION_R_NUMELEM; t++) {
            if (invalid_var(ekfConfiguration.R[t])) {
                error = true;
            }
        }

        RevoSettingsGet(&revoSettings);

        // Reinitialization of the EKF is not desired during flight.
        // It will be delayed until the board is disarmed by raising the error flag.
        // We will not prevent the initial initialization though, since the board could be in always armed mode.
        if (flightStatus.Armed == FLIGHTSTATUS_ARMED_ARMED && !initialization_required) {
            error = true;
        }

        if (error) {
            variance_error = true;
        } else {
            // trigger reinitialization - possibly with new algorithm
            running_algorithm = revoSettings.FusionAlgorithm;
            variance_error    = false;
            initialization_required = true;
        }
    }
    if (ev == NULL || ev->obj == HomeLocationHandle()) {
        HomeLocationGet(&homeLocation);
        // Compute matrix to convert deltaLLA to NED
        float lat, alt;
        lat  = DEG2RAD(homeLocation.Latitude / 10.0e6f);
        alt  = homeLocation.Altitude;

        T[0] = alt + 6.378137E6f;
        T[1] = cosf(lat) * (alt + 6.378137E6f);
        T[2] = -1.0f;

        // TODO: convert positionState to new reference frame and gracefully update EKF state!
        // needed for long range flights where the reference coordinate is adjusted in flight
    }
    if (ev == NULL || ev->obj == AttitudeSettingsHandle()) {
        AttitudeSettingsGet(&attitudeSettings);

        // Calculate accel filter alpha, in the same way as for gyro data in stabilization module.
        const float fakeDt = 0.0015f;
        if (attitudeSettings.AccelTau < 0.0001f) {
            accel_alpha = 0; // not trusting this to resolve to 0
            accel_filter_enabled = false;
        } else {
            accel_alpha = expf(-fakeDt / attitudeSettings.AccelTau);
            accel_filter_enabled = true;
        }
    }
}

/**
 * Perform an update of the @ref MagBias based on
 * Magmeter Offset Cancellation: Theory and Implementation,
 * revisited William Premerlani, October 14, 2011
 */
static void magOffsetEstimation(MagSensorData *mag)
{
   #if 0
    // Constants, to possibly go into a UAVO
    static const float MIN_NORM_DIFFERENCE = 50;

    static float B2[3] = { 0, 0, 0 };

    MagBiasData magBias;
    MagBiasGet(&magBias);

    // Remove the current estimate of the bias
    mag->x -= magBias.x;
    mag->y -= magBias.y;
    mag->z -= magBias.z;

    // First call
    if (B2[0] == 0 && B2[1] == 0 && B2[2] == 0) {
        B2[0] = mag->x;
        B2[1] = mag->y;
        B2[2] = mag->z;
        return;
    }

    float B1[3]     = { mag->x, mag->y, mag->z };
    float norm_diff = sqrtf(powf(B2[0] - B1[0], 2) + powf(B2[1] - B1[1], 2) + powf(B2[2] - B1[2], 2));
    if (norm_diff > MIN_NORM_DIFFERENCE) {
        float norm_b1    = sqrtf(B1[0] * B1[0] + B1[1] * B1[1] + B1[2] * B1[2]);
        float norm_b2    = sqrtf(B2[0] * B2[0] + B2[1] * B2[1] + B2[2] * B2[2]);
        float scale      = cal.MagBiasNullingRate * (norm_b2 - norm_b1) / norm_diff;
        float b_error[3] = { (B2[0] - B1[0]) * scale, (B2[1] - B1[1]) * scale, (B2[2] - B1[2]) * scale };

        magBias.x += b_error[0];
        magBias.y += b_error[1];
        magBias.z += b_error[2];

        MagBiasSet(&magBias);

        // Store this value to compare against next update
        B2[0] = B1[0]; B2[1] = B1[1]; B2[2] = B1[2];
    }
   #else // if 0
    static float magBias[3] = { 0 };

    // Remove the current estimate of the bias
    mag->x -= magBias[0];
    mag->y -= magBias[1];
    mag->z -= magBias[2];

    AttitudeStateData attitude;
    AttitudeStateGet(&attitude);

    const float Rxy  = sqrtf(homeLocation.Be[0] * homeLocation.Be[0] + homeLocation.Be[1] * homeLocation.Be[1]);
    const float Rz   = homeLocation.Be[2];

    const float rate = revoCalibration.MagBiasNullingRate;
    float Rot[3][3];
    float B_e[3];
    float xy[2];
    float delta[3];

    // Get the rotation matrix
    Quaternion2R(&attitude.q1, Rot);

    // Rotate the mag into the NED frame
    B_e[0] = Rot[0][0] * mag->x + Rot[1][0] * mag->y + Rot[2][0] * mag->z;
    B_e[1] = Rot[0][1] * mag->x + Rot[1][1] * mag->y + Rot[2][1] * mag->z;
    B_e[2] = Rot[0][2] * mag->x + Rot[1][2] * mag->y + Rot[2][2] * mag->z;

    float cy = cosf(DEG2RAD(attitude.Yaw));
    float sy = sinf(DEG2RAD(attitude.Yaw));

    xy[0] = cy * B_e[0] + sy * B_e[1];
    xy[1] = -sy * B_e[0] + cy * B_e[1];

    float xy_norm = sqrtf(xy[0] * xy[0] + xy[1] * xy[1]);

    delta[0] = -rate * (xy[0] / xy_norm * Rxy - xy[0]);
    delta[1] = -rate * (xy[1] / xy_norm * Rxy - xy[1]);
    delta[2] = -rate * (Rz - B_e[2]);

    if (!isnan(delta[0]) && !isinf(delta[0]) &&
        !isnan(delta[1]) && !isinf(delta[1]) &&
        !isnan(delta[2]) && !isinf(delta[2])) {
        magBias[0] += delta[0];
        magBias[1] += delta[1];
        magBias[2] += delta[2];
    }
    #endif // if 0
}


/**
 * @}
 * @}
 */