The two state auxiliary EKF has been replaced with a single state filter that only estimates terrain offset. The new filter fuses a optical flow line of sight rate scalar (length of the optical flow LOS rate vector) which provides a terrain offset estimate that is less affected by yaw errors.
Estimation of focal length scale factor error in flight wasn't accurate enough and will be replaced with a pre-flight intrinsic sensor calibration procedure as the scale factor error does not change over time provided the lens assembly is not adjusted.
AP_NavEKF: Remove unwanted printf
The terrain offset solution status is usable for a short period of time without state updates so a timeout has been added which prevents the rapid changes in solution status due to short duration sensor read errors.
This fixes a bug which could have caused the realative position status to be incorrectly reported under some conditions and also caused a compiler warning message. the logic used to report the filter solution status has been broken down into smaller, easier to understand statements.
Non user adjustable parameters are now declared as 'const' in the header.
The _ prefix has been removed from non user adjustable tuning parameters.
We use a function call rather than a constructor to initialise variables because it enables the filter to be re-started in flight if necessary.
For consistency some signed integer type declarations have been changed to unsigned where appropriate.
The decision to switch to constant velocity mode during optical flow operation and te decision to switch back were previously being made in two different places in code. Both decisions are now made in the one place which makes the code easier to follow and maintain.
Move velocity store out of optical flow to velocity and position fusion control as it is a velocity fusion function.
Always clear the previous mode status
This patch makes the reporting of an absolute position solution less abbiguaous and ensures that relative position is always true if absolute position is true
Because synthetic position measurements at 0,0 are fused during position hold mode, a position timeout cannot be used as the only means of detecting position solution status.
Required to prevent acquisition of GPS mid flight causing unwanted change in position and velocity
A distinction has been mad between the arm and disarm transition and the decision to use position hold mode (formerly static mode)
When regaining GPS after a timeout, an offset is applied when fusing GPS velocity so that GPS velocity and position data as fused by the EKF is kinematically consistent.
This velocity offset is also accounted for when fusing air data so that wind estimates are not corrupted when the GPS position offset is being pulle back to zero.
The intended behaviour is that the EKF position will be pulled back to the GPS position at a rate of 5m/s for planes and 1 m/s for copters. This avoids large deviations in trajectory when GPS is regained.
When on the ground it is likely the flow sensor will be returning data that does not meet the minimum quality requirements selected.
The previous check was for the presence of valid data. This has been loosened to look for the presence of data.
When the vehicle becomes airborne, the quality of flow data normally improves as the image comes into focus.
1) Un-used public methods to report height and position drifting have been removed
2) A time-out has been added to the airspeed innovation consistency check so that if we are relying on airspeed to constrain velocity drift, a filter divergence or other fault that causes the airspeed to be continually rejected will trigger a change in health status.
3) A timeout of velocity, position or height measurements does not cause a filter fault to be reported. Timeouts can be due to sensor errors and do not necessarily indicate that the filter has failed.
4) Time-outs of various measurements are used to present a consolidated bitmask which inidicates which parts of the solution can be used, eg attitude, height, velocity, relative position, absolute position, etc.
The time required for GPS to be lost or rejected before vehicles with airspeed sensors either reset to GPS or invoke the zero side-slip assumption to constrain drift has been reduced from 15 to 10 seconds
A duplicate zeroing of the GPS position offset has been removed
If the vehicle is a non hovering vehicle (eg a plane) then the speed at which the GPS offset is pulled back to zero after a reset is increased from 1 to 3 m/s
This also improves recovery from bad inertial data for planes
Doing this can cause large height and height rate errors if large GPS velocity errors cause the GPS tn be rejected for long enough to cause a timeout and reset of states.
This reverts commit 13df6fb1c9.
This patch was a result of an incorrect merge of the optflow branch
into master. It reintroduced the bug fixed by this commit:
AP_NavEKF : Fix bug in reset of GPS glitch offset
8aa267f75f
This provides the calling vehicle software the abiity to request the EKF to not use GPS.
An integer is returned that indicates the type of operation available:
0 = request rejected (request will only be accepted if the EKF is in static mode, eg pre-armed)
1 = request accepted, attitude, vertical velocity and position estimates available
2 = request accepted, attitude, height rate, height, horizontal velocity and relative position estimates available
This takes into account the inter-sampling delay between the flow driver and the APM software which depends on the rate at which the dirver is being checked. This is 50Hz for plane and rover, and 200Hz for Copter.
This patch reduces the level of 5Hz and 10Hz 'pulsing' heard in motors due to GPS and altimeter fusion which cause a small 5Hz and 10Hz ripple on the output under some conditions. Attitude, velocity and position state corrections from GPS, altimeter and magnetometer measurements are applied incrementally in the interval from receiving the measurement to the predicted time of receipt of the next measurement. Averaging of attitude state corrections is not performed during periods of rapid rotation.
Time stamps are now explicitly initialised to the current IMU time to avoid unwanted activation of timeout logic on filter start and the various calls to the hal.scheduler->millis() object have been consolidated.
Unused variables have been removed
If the inertial solution velocity or position needs to be reset to the GPS, the stored state history for the corresponding states should also be reset.
Otherwise the next GPS measurement will be compared to an invalid previous state and will be rejected.
The position state should be reset to a GPS position corrected for velocity and measurement latency. This will make a noticeable difference for high speed flight vehicles, eg 11m at 50m/s
Re-initialisation of the magnetic field states and yaw angle is now only performed a maximum of two times after start-up.
Once when coming out of static modefor the first time (first arm event)
Again (for copter only) when the altitude gain above the arming altitude exceeds 1.5m
this prevents magnetic interference present at arming (eg arming on a metal roof)from corrupting the magnetic field states enough to cause bad heading errors and toilet bowling on copter
(Plane Only) If the yaw and GPS heading disagree by more than 45 degrees on takeoff, then the magnetometer is declared as failed. The heading is then reset based on the difference between GPS ground track and stgate velocity vector.
Magnetometer fusion uses corrected data and bias states are initialised to zero. This allows the compass to be switched in flight.
For persistent compass errors that trigger a timeout, the compass is not permanently failed, however for non-forward fly vehicles the compass weighting is reduced.
The GPS glitch offset was being zeroed during position resets. This caused the filter to reject subsequent GPS measurements if the GPS error persisted long enough to invoke a timeout and a position reset.
GPS measurement variance is doubled if only 5 satellites, and quadrupled if 4 or less.
The GPS glitch rejection thresholds remain the same
This will reduce the impact of GPS glitches on attitude.
The uncertainty in acceleration is currently only scaled using horizontal accelerations, however during vertical plane aerobatics and high g pullups, misalignment in angles can cause significant horizontal acceleration error.
The scaling now uses the 3D acceleration vector length to better work during vertical plane high g maneouvres.
This error was found during flight testing with 8g pullups
If the inertial solution velocity or position needs to be reset to the GPS or baro, the stored state history for the corresponding states should also be reset.
Otherwise the next GPS or baro measurement will be compared to an invalid previous state and will be rejected. This is particularly a problem if IMU saturation or timeout has occurred because the previous states could be out by a large amount
The position state should be reset to a GPS position corrected for velocity and measurement latency. This will make a noticeable difference for high speed flight vehicles, eg 11m at 50m/s.
Flying aerobatics with Trad Heli has shown that the divergence check can be false triggered when large magnetometer errors and GPS dropouts are present.
This can also happen with multi rotors if large yaw rates are present.
This was an unintended consequence of the ekfsmoothing patch which improved filter stability during high rate manoeuvres, but made the divergence test more sensitive.
This patch reduces the level of 5Hz and 10Hz 'pulsing' heard in motors due to GPS and altimeter fusion which cause a small 5Hz and 10Hz ripple on the output under some conditions. Attitude, velocity and position state corrections from GPS, altimeter and magnetometer measurements are applied incrementally in the interval from receiving the measurement to the predicted time of receipt of the next measurement. Averaging of attitude state corrections is not performed during periods of rapid rotation.
Time stamps are now explicitly initialised to the current IMU time to avoid unwanted activation of timeout logic on filter start and various calls to the hal.scheduler->millis() object have been consolidated.
This allows a compass that has been declared failed, possibly because of
external disturbances (eg movement of hatches, proximity of tools, etc)
to be given a second chance when the vehicle is armed.
If the vehicle can fly without a compass (a fly forward vehicle)
then if the compass times out (large errors for more than 10 seconds,
then it will be declared permanently failed and will not be
used until the filter is reset
Analysis of copter logs has shown cases with a healthy EKF where spikes in EKF4.DS of up to 25% of the threshold have occurred.
A value of closer to 10% for normal operation is preferred.
Don't do bias estimation if tilted by more than 60 degrees to prevent scale
factor errors affecting result unnecessarily.
Prevent Kalman gain from having the wrong sign due to numerical errors
associated with small process noise values.
Allow smaller Z accel bias process noise values to be set
Bring Plane glitch protection thresholds into alignment with copter and
rover
Slight increase in accelerometer bias process noise to prevent bias
estimate divergence into limits (Rover and Plane only as Copter does not
seem respond as well to this change)
effective increase in threshold on divergence test to allow increased
margin for bad GPS velocities
Divergence is now detected by looking for very large changes in the gyro
bias. This will cause the filter to be reset and declared unhealthy for
10 seconds.
Don't reset filterDiverged status immediately during reset
Set filterDiverged true if covariance blows up
Add fault status reporting
This will enable in-flight magnetometer calibration to be inhibited unconditionally,
This is required for long balloon carriage flights where ground speed can be high
enough to put it into in-air state, but with very poor observability of magnetic field
states causing bad state estimates and heading offsets to develop over time.
The covariance matrix no longer has rows and columns artificially zeroed when in static
mode. Instead booleans indicating whether wind or magentic field state estimation is
active are used to:
a) Set the process noise on these states to zero to stop their variances from increasing
unchecked when not being updated, and
b) Turn off updates for these states when measurement fusion is being performed.
This reduces the likelihood of a badly conditioned covariance matrix forming
during static mode operation.
A filter divergence check has also been added that will declare the filter unhealthy if
position, velocity and magnetic field observations are all failing their innovation
consistency checks. This unhealthy status will persist for 10 seconds after the
condition clears.
AP_NavEKF: Remove unnecessary zeroing of wind covariances
A magnetometer axis that fails the innovation consistency check will cause
all axes not to be used. If this condition continues for 10 seconds, a
magnetometer timeout condition will be declared. When the timeout has
occurred, if it is not a fly forward vehicle, then individual channels
will be used again, but with a reduced weighting.
this will cope better with users with low roll/pitch gains, to ensure
they get enough control on takeoff
Pair-Programmed-With: Paul Riseborough <p_riseborough@live.com.au>
Flight testing with windup turns has shown that the position gate threshold
can be tripped with good GPS data causing position jerks. This increases the
initial GPS glitch rejection threshold to effectively 5m when using the
default POSNE_NOISE value of 0.5m.
This patch also cleans up the logic associated with use of the synthetic
sideslip measurement so that it can never be used for a non fly-forward
vehicle type
This patch adds functionality that initialises the wind-speed vector to the
reciprocal of the ground speed vector, and scaled to 6 m/s. On average this gives
a better initial wind velocity estimate on launch by assuming:
a) launch will be into wind
b) wind speed is equal to global average
It also helps prevent a headwind causing initial underestimation of airspeed
causing high autopilot gains and limit cycles on climb-out, until first
turn when the EKF is able to estimate the wind.
When using GPS after previously rejecting it, the GPS position will
always be offset if outside the specified glitch radius. This was the
original intent of the design and makes handling of glitches smoother.
It has been tested on replay using glitchy flight data
Aliasing can causes the bias estimate to fluctuate very rapidly as it tries
to keep up, which degrades the benefit of switching between
accelerometers to avoid aliasing.
This patch give a much more stable bias estimate during aliasing, and
allows the bias to adapt at a maximum rate of 1.0 m/s2 in 50 seconds
This adds new functionality to the detection and compensation of GPS
glitches:
1) A maximum allowable innovation is calculated using the GPS noise
parameter multiplied by the gate, with an additional component allowing
for growth in position uncertainty due to acceleration error since
the last valid measurement
2) Includes per vehicle type values for the acceleration error limit
3) If the innovation length exceeds the maximum allowable, no fusion occurs
4) If no fusion has occurred for long enough such that the position uncertainty
exceeds the maximum set by a per vehicle parameter or a maximum time, an offset
is applied to the GPS data to so that it matches the value predicted by the filter
5) The offset is never allowed to be bigger than 100m
6) The offset is decayed to zero at a rate of 1.0 m/s to allow GPS jumps to
be accommodated gradually
7) The default velocity innovation gate has been tightened up for copter and rover
8) The variance data logging output has been updated to make it more useful
This patch reduces the maximum acceptable GPS jump from approximately 16 to 8 metres
This will provide copters with more protection for close in loiter situations
This path reduces duplicated code, eliminates unused variables and
causes the earth magnetic field states to be reset when exiting static mode
which will occur every time copter is armed. This enables copters to be
powered on and initialised inside vehicles or houses, without bad earth
field values affecting flight.
Constraining variances to a minimum value of 1e-9 was causing problems
with gyro bias and angular accuracy in noisy GPS environments.
Because the constraint is applied after every covariance prediction
and correction, a lower value of 0 is more appropriate.
priseborough
Randy Mackay
Andrew Tridgell
Jonathan Challinger
Andrew Tridgell
Emile Castelnuovo