2012-07-28 02:17:38 -03:00
|
|
|
|
/*
|
|
|
|
|
AP_AHRS_MPU6000.cpp
|
|
|
|
|
|
|
|
|
|
AHRS system using MPU6000's internal calculations
|
|
|
|
|
|
|
|
|
|
Adapted for the general ArduPilot AHRS interface by Andrew Tridgell
|
|
|
|
|
|
|
|
|
|
This library is free software; you can redistribute it and/or
|
|
|
|
|
modify it under the terms of the GNU Lesser General Public License
|
|
|
|
|
as published by the Free Software Foundation; either version 2.1
|
|
|
|
|
of the License, or (at your option) any later version.
|
|
|
|
|
*/
|
|
|
|
|
#include <FastSerial.h>
|
|
|
|
|
#include <AP_AHRS.h>
|
|
|
|
|
|
|
|
|
|
// this is the speed in cm/s above which we first get a yaw lock with
|
|
|
|
|
// the GPS
|
|
|
|
|
#define GPS_SPEED_MIN 300
|
|
|
|
|
|
|
|
|
|
// this is the speed in cm/s at which we stop using drift correction
|
|
|
|
|
// from the GPS and wait for the ground speed to get above GPS_SPEED_MIN
|
|
|
|
|
#define GPS_SPEED_RESET 100
|
|
|
|
|
|
|
|
|
|
// the limit (in degrees/second) beyond which we stop integrating
|
|
|
|
|
// omega_I. At larger spin rates the DCM PI controller can get 'dizzy'
|
|
|
|
|
// which results in false gyro drift. See
|
|
|
|
|
// http://gentlenav.googlecode.com/files/fastRotations.pdf
|
|
|
|
|
#define SPIN_RATE_LIMIT 20
|
|
|
|
|
|
|
|
|
|
// table of user settable parameters
|
|
|
|
|
const AP_Param::GroupInfo AP_AHRS_MPU6000::var_info[] PROGMEM = {
|
|
|
|
|
// @Param: AHRS_YAW_P
|
|
|
|
|
// @DisplayName: Yaw P
|
|
|
|
|
// @Description: This controls the weight the compass has on the overall heading
|
|
|
|
|
// @Range: 0 .4
|
|
|
|
|
// @Increment: .01
|
2012-08-06 22:01:44 -03:00
|
|
|
|
AP_GROUPINFO("YAW_P", 0, AP_AHRS_MPU6000, _kp_yaw, 0.4),
|
2012-07-28 02:17:38 -03:00
|
|
|
|
|
|
|
|
|
// @Param: AHRS_RP_P
|
|
|
|
|
// @DisplayName: AHRS RP_P
|
|
|
|
|
// @Description: This controls how fast the accelerometers correct the attitude
|
|
|
|
|
// @Range: 0 .4
|
|
|
|
|
// @Increment: .01
|
2012-08-06 22:01:44 -03:00
|
|
|
|
AP_GROUPINFO("RP_P", 1, AP_AHRS_MPU6000, _kp, 0.4),
|
2012-07-28 02:17:38 -03:00
|
|
|
|
AP_GROUPEND
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
AP_AHRS_MPU6000::init()
|
|
|
|
|
{
|
|
|
|
|
_mpu6000->dmp_init();
|
|
|
|
|
push_gains_to_dmp();
|
|
|
|
|
push_offsets_to_ins();
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
// run a full MPU6000 update round
|
|
|
|
|
void
|
|
|
|
|
AP_AHRS_MPU6000::update(void)
|
|
|
|
|
{
|
|
|
|
|
float delta_t;
|
|
|
|
|
|
|
|
|
|
// tell the IMU to grab some data. is this necessary?
|
|
|
|
|
_imu->update();
|
|
|
|
|
|
|
|
|
|
// ask the IMU how much time this sensor reading represents
|
|
|
|
|
delta_t = _imu->get_delta_time();
|
|
|
|
|
|
|
|
|
|
// convert the quaternions into a DCM matrix
|
|
|
|
|
_mpu6000->quaternion.rotation_matrix(_dcm_matrix);
|
|
|
|
|
|
|
|
|
|
// we run the gyro bias correction using gravity vector algorithm
|
|
|
|
|
drift_correction(delta_t);
|
|
|
|
|
|
|
|
|
|
// Calculate pitch, roll, yaw for stabilization and navigation
|
|
|
|
|
euler_angles();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// wrap_PI - ensure an angle (expressed in radians) is between -PI and PI
|
|
|
|
|
// TO-DO: should remove and replace with more standard functions
|
|
|
|
|
float AP_AHRS_MPU6000::wrap_PI(float angle_in_radians)
|
|
|
|
|
{
|
|
|
|
|
if( angle_in_radians > M_PI ){
|
|
|
|
|
return(angle_in_radians - 2*M_PI);
|
|
|
|
|
}
|
|
|
|
|
else if( angle_in_radians < -M_PI ){
|
|
|
|
|
return(angle_in_radians + 2*M_PI);
|
|
|
|
|
}
|
|
|
|
|
else{
|
|
|
|
|
return(angle_in_radians);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Function to correct the gyroX and gyroY bias (roll and pitch) using the gravity vector from accelerometers
|
|
|
|
|
// We use the internal chip axis definition to make the bias correction because both sensors outputs (gyros and accels)
|
|
|
|
|
// are in chip axis definition
|
|
|
|
|
void AP_AHRS_MPU6000::drift_correction( float deltat )
|
|
|
|
|
{
|
|
|
|
|
float errorRollPitch[2];
|
|
|
|
|
|
|
|
|
|
// Get current values for gyros
|
|
|
|
|
_accel_vector = _imu->get_accel();
|
|
|
|
|
|
|
|
|
|
// We take the accelerometer readings and cumulate to average them and obtain the gravity vector
|
|
|
|
|
_accel_filtered += _accel_vector;
|
|
|
|
|
_accel_filtered_samples++;
|
|
|
|
|
|
|
|
|
|
_gyro_bias_from_gravity_counter++;
|
|
|
|
|
// We make the bias calculation and correction at a lower rate (GYRO_BIAS_FROM_GRAVITY_RATE)
|
|
|
|
|
if( _gyro_bias_from_gravity_counter == GYRO_BIAS_FROM_GRAVITY_RATE ){
|
|
|
|
|
_gyro_bias_from_gravity_counter = 0;
|
|
|
|
|
|
|
|
|
|
_accel_filtered /= _accel_filtered_samples; // average
|
|
|
|
|
|
|
|
|
|
// Adjust ground reference : Accel Cross Gravity to obtain the error between gravity from accels and gravity from attitude solution
|
|
|
|
|
// errorRollPitch are in Accel LSB units
|
|
|
|
|
errorRollPitch[0] = _accel_filtered.y * _dcm_matrix.c.z + _accel_filtered.z * _dcm_matrix.c.x;
|
|
|
|
|
errorRollPitch[1] = -_accel_filtered.z * _dcm_matrix.c.y - _accel_filtered.x * _dcm_matrix.c.z;
|
|
|
|
|
|
|
|
|
|
errorRollPitch[0] *= deltat * 1000;
|
|
|
|
|
errorRollPitch[1] *= deltat * 1000;
|
|
|
|
|
|
|
|
|
|
// we limit to maximum gyro drift rate on each axis
|
|
|
|
|
float drift_limit = _mpu6000->get_gyro_drift_rate() * deltat / _gyro_bias_from_gravity_gain; //0.65*0.04 / 0.005 = 5.2
|
|
|
|
|
errorRollPitch[0] = constrain(errorRollPitch[0], -drift_limit, drift_limit);
|
|
|
|
|
errorRollPitch[1] = constrain(errorRollPitch[1], -drift_limit, drift_limit);
|
|
|
|
|
|
|
|
|
|
// We correct gyroX and gyroY bias using the error vector
|
|
|
|
|
_gyro_bias[0] += errorRollPitch[0]*_gyro_bias_from_gravity_gain;
|
|
|
|
|
_gyro_bias[1] += errorRollPitch[1]*_gyro_bias_from_gravity_gain;
|
|
|
|
|
|
|
|
|
|
// TO-DO: fix this. Currently it makes the roll and pitch drift more!
|
|
|
|
|
// If bias values are greater than 1 LSB we update the hardware offset registers
|
|
|
|
|
if( fabs(_gyro_bias[0])>1.0 ){
|
|
|
|
|
//_mpu6000->set_gyro_offsets(-1*(int)_gyro_bias[0],0,0);
|
|
|
|
|
//_mpu6000->set_gyro_offsets(0,-1*(int)_gyro_bias[0],0);
|
|
|
|
|
//_gyro_bias[0] -= (int)_gyro_bias[0]; // we remove the part that we have already corrected on registers...
|
|
|
|
|
}
|
|
|
|
|
if (fabs(_gyro_bias[1])>1.0){
|
|
|
|
|
//_mpu6000->set_gyro_offsets(-1*(int)_gyro_bias[1],0,0);
|
|
|
|
|
//_gyro_bias[1] -= (int)_gyro_bias[1];
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Reset the accelerometer variables
|
|
|
|
|
_accel_filtered.x = 0;
|
|
|
|
|
_accel_filtered.y = 0;
|
|
|
|
|
_accel_filtered.z = 0;
|
|
|
|
|
_accel_filtered_samples=0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// correct the yaw
|
|
|
|
|
drift_correction_yaw();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
reset the DCM matrix and omega. Used on ground start, and on
|
|
|
|
|
extreme errors in the matrix
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
AP_AHRS_MPU6000::reset(bool recover_eulers)
|
|
|
|
|
{
|
|
|
|
|
// if the caller wants us to try to recover to the current
|
|
|
|
|
// attitude then calculate the dcm matrix from the current
|
|
|
|
|
// roll/pitch/yaw values
|
|
|
|
|
if (recover_eulers && !isnan(roll) && !isnan(pitch) && !isnan(yaw)) {
|
|
|
|
|
_dcm_matrix.from_euler(roll, pitch, yaw);
|
|
|
|
|
} else {
|
|
|
|
|
// otherwise make it flat
|
|
|
|
|
_dcm_matrix.from_euler(0, 0, 0);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// push offsets down from IMU to INS (required so MPU6000 can perform it's own attitude estimation)
|
|
|
|
|
void
|
|
|
|
|
AP_AHRS_MPU6000::push_offsets_to_ins()
|
|
|
|
|
{
|
|
|
|
|
// push down gyro offsets (TO-DO: why are x and y offsets are reversed?!)
|
|
|
|
|
_mpu6000->set_gyro_offsets_scaled(((AP_IMU_INS*)_imu)->gy(),((AP_IMU_INS*)_imu)->gx(),((AP_IMU_INS*)_imu)->gz());
|
|
|
|
|
((AP_IMU_INS*)_imu)->gx(0);
|
|
|
|
|
((AP_IMU_INS*)_imu)->gy(0);
|
|
|
|
|
((AP_IMU_INS*)_imu)->gz(0);
|
|
|
|
|
|
|
|
|
|
// push down accelerometer offsets (TO-DO: why are x and y offsets are reversed?!)
|
|
|
|
|
_mpu6000->set_accel_offsets_scaled(((AP_IMU_INS*)_imu)->ay(), ((AP_IMU_INS*)_imu)->ax(), ((AP_IMU_INS*)_imu)->az());
|
|
|
|
|
//((AP_IMU_INS*)_imu)->ax(0);
|
|
|
|
|
//((AP_IMU_INS*)_imu)->ay(0);
|
|
|
|
|
//((AP_IMU_INS*)_imu)->az(0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
AP_AHRS_MPU6000::push_gains_to_dmp()
|
|
|
|
|
{
|
|
|
|
|
uint8_t gain;
|
|
|
|
|
if( _kp.get() >= 1.0 ) {
|
|
|
|
|
gain = 0xFF;
|
|
|
|
|
}else if( _kp.get() <= 0.0 ) {
|
|
|
|
|
gain = 0x00;
|
|
|
|
|
}else{
|
|
|
|
|
gain = (uint8_t)((float)0xFF * _kp.get());
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
_mpu6000->dmp_set_sensor_fusion_accel_gain(gain);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// produce a yaw error value. The returned value is proportional
|
|
|
|
|
// to sin() of the current heading error in earth frame
|
|
|
|
|
float
|
|
|
|
|
AP_AHRS_MPU6000::yaw_error_compass(void)
|
|
|
|
|
{
|
|
|
|
|
Vector3f mag = Vector3f(_compass->mag_x, _compass->mag_y, _compass->mag_z);
|
|
|
|
|
// get the mag vector in the earth frame
|
|
|
|
|
Vector3f rb = _dcm_matrix * mag;
|
|
|
|
|
|
|
|
|
|
rb.normalize();
|
|
|
|
|
if (rb.is_inf()) {
|
|
|
|
|
// not a valid vector
|
|
|
|
|
return 0.0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// get the earths magnetic field (only X and Y components needed)
|
|
|
|
|
Vector3f mag_earth = Vector3f(cos(_compass->get_declination()),
|
|
|
|
|
sin(_compass->get_declination()), 0);
|
|
|
|
|
|
|
|
|
|
// calculate the error term in earth frame
|
|
|
|
|
Vector3f error = rb % mag_earth;
|
|
|
|
|
|
|
|
|
|
return error.z;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
//
|
|
|
|
|
// drift_correction_yaw - yaw drift correction using the compass
|
|
|
|
|
//
|
|
|
|
|
void
|
|
|
|
|
AP_AHRS_MPU6000::drift_correction_yaw(void)
|
|
|
|
|
{
|
|
|
|
|
static float yaw_last_uncorrected = HEADING_UNKNOWN;
|
|
|
|
|
static float yaw_corrected = HEADING_UNKNOWN;
|
|
|
|
|
float yaw_delta = 0;
|
|
|
|
|
bool new_value = false;
|
|
|
|
|
float yaw_error;
|
|
|
|
|
static float heading;
|
|
|
|
|
|
|
|
|
|
// we assume the DCM matrix is completely uncorrect for yaw
|
|
|
|
|
// retrieve how much heading has changed according to non-corrected dcm
|
|
|
|
|
if( yaw_last_uncorrected != HEADING_UNKNOWN ) {
|
|
|
|
|
yaw_delta = wrap_PI(yaw - yaw_last_uncorrected); // the change in heading according to the gyros only since the last interation
|
|
|
|
|
yaw_last_uncorrected = yaw;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// initialise yaw_corrected (if necessary)
|
|
|
|
|
if( yaw_corrected != HEADING_UNKNOWN ) {
|
|
|
|
|
yaw_corrected = yaw;
|
|
|
|
|
}else{
|
|
|
|
|
yaw_corrected = wrap_PI(yaw_corrected + yaw_delta); // best guess of current yaw is previous iterations corrected yaw + yaw change from gyro
|
|
|
|
|
_dcm_matrix.from_euler(roll, pitch, yaw_corrected); // rebuild dcm matrix with best guess at current yaw
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// if we have new compass data
|
|
|
|
|
if( _compass && _compass->use_for_yaw() ) {
|
|
|
|
|
if (_compass->last_update != _compass_last_update) {
|
|
|
|
|
_compass_last_update = _compass->last_update;
|
|
|
|
|
heading = _compass->calculate_heading(_dcm_matrix);
|
|
|
|
|
if( !_have_initial_yaw ) {
|
|
|
|
|
yaw_corrected = heading;
|
|
|
|
|
_have_initial_yaw = true;
|
|
|
|
|
_dcm_matrix.from_euler(roll, pitch, yaw_corrected); // rebuild dcm matrix with best guess at current yaw
|
|
|
|
|
}
|
|
|
|
|
new_value = true;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// perform the yaw correction
|
|
|
|
|
if( new_value ){
|
|
|
|
|
yaw_error = yaw_error_compass();
|
|
|
|
|
// the yaw error is a vector in earth frame
|
|
|
|
|
//Vector3f error = Vector3f(0,0, yaw_error);
|
|
|
|
|
|
|
|
|
|
// convert the error vector to body frame
|
|
|
|
|
//error = _dcm_matrix.mul_transpose(error);
|
|
|
|
|
|
|
|
|
|
// Update the differential component to reduce the error (it<69>s like a P control)
|
|
|
|
|
yaw_corrected += wrap_PI(yaw_error * _kp_yaw.get()); // probably completely wrong
|
|
|
|
|
|
|
|
|
|
// rebuild the dcm matrix yet again
|
|
|
|
|
_dcm_matrix.from_euler(roll, pitch, yaw_corrected);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// calculate the euler angles which will be used for high level
|
|
|
|
|
// navigation control
|
|
|
|
|
void
|
|
|
|
|
AP_AHRS_MPU6000::euler_angles(void)
|
|
|
|
|
{
|
|
|
|
|
_dcm_matrix.to_euler(&roll, &pitch, &yaw);
|
|
|
|
|
//quaternion.to_euler(&roll, &pitch, &yaw); // cannot use this because the quaternion is not correct for yaw drift
|
|
|
|
|
|
|
|
|
|
roll_sensor = degrees(roll) * 100;
|
|
|
|
|
pitch_sensor = degrees(pitch) * 100;
|
|
|
|
|
yaw_sensor = degrees(yaw) * 100;
|
|
|
|
|
|
|
|
|
|
if (yaw_sensor < 0)
|
|
|
|
|
yaw_sensor += 36000;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* reporting of DCM state for MAVLink */
|
|
|
|
|
|
|
|
|
|
// average error_roll_pitch since last call
|
|
|
|
|
float AP_AHRS_MPU6000::get_error_rp(void)
|
|
|
|
|
{
|
|
|
|
|
// not yet supported with DMP
|
|
|
|
|
return 0.0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// average error_yaw since last call
|
|
|
|
|
float AP_AHRS_MPU6000::get_error_yaw(void)
|
|
|
|
|
{
|
|
|
|
|
// not yet supported with DMP
|
|
|
|
|
return 0.0;
|
|
|
|
|
}
|