temperature_compensation module #12601

dagar · 2019-08-01T20:35:26Z

This is a new module for temperature compensation that's currently part of the sensors and events modules.

Moving this to a separate module means it can be trivially disabled when not used (saving ~ 16 kB of flash), and is architecturally consistent with removing the sensors hub as a bottle neck for running multiple estimators.

dagar · 2019-08-01T20:35:56Z

Flash savings if disabled from a build (eg things get desperate in px4_fmu-v2).

dagar · 2019-08-01T20:51:04Z

src/modules/sensors/voted_sensors_update.cpp

@@ -730,7 +659,6 @@ void VotedSensorsUpdate::gyro_poll(struct sensor_combined_s &raw)
 		if (_gyro.last_best_vote != best_index) {
 			_gyro.last_best_vote = (uint8_t)best_index;
 			_corrections.selected_gyro_instance = (uint8_t)best_index;


This is wrong.

The selected_gyro_instance needs to be published.

Hmm... I think we need to split these topics
I'd like to rename sensor_correction to sensor_thermal_corrections and move

uint8_t selected_gyro_instance; uint8_t selected_accel_instance; uint8_t selected_baro_instance;

into their own topic (active_sensors? select_sensors?). Does this make sense?

The *_mapping is just for applying the thermal corrections to the correct senor...right?

I believe sensor_correction could move entirely to temperature_compensation and the selected instance (and mapping) could go away entirely.

Then in the sensors module sensor_selection could also contain the ORB instance number.

julianoes · 2019-08-05T12:13:35Z

@dagar please request reviews when it's ready.

dagar · 2019-09-30T02:08:50Z

This is working now (at least on the bench). @dakejahl could you take another pass?

nsh> temperature_compensation status
INFO  [temperature_compensation] Temperature Compensation:
INFO  [temperature_compensation]  gyro: enabled: 1
INFO  [temperature_compensation]   using device ID 3672842 for topic instance 0
INFO  [temperature_compensation]   using device ID 2360330 for topic instance 1
INFO  [temperature_compensation]  accel: enabled: 1
INFO  [temperature_compensation]   using device ID 3607306 for topic instance 0
INFO  [temperature_compensation]   using device ID 1442826 for topic instance 1
INFO  [temperature_compensation]  baro: enabled: 1
INFO  [temperature_compensation]   using device ID 3998482 for topic instance 0

nsh> sensors status
INFO  [sensors] gyro status:
INFO  [ecl/validation] validator: best: 0, prev best: 0, failsafe: NO (0 events)
INFO  [ecl/validation] sensor #0, prio: 100, state: OK
INFO  [ecl/validation]  val:  -0.0046, lp:  -0.0010 mean dev:  -0.0000 RMS:   0.0028 conf:   1.0000
INFO  [ecl/validation]  val:   0.0035, lp:   0.0012 mean dev:  -0.0001 RMS:   0.0015 conf:   1.0000
INFO  [ecl/validation]  val:   0.0020, lp:   0.0022 mean dev:   0.0001 RMS:   0.0010 conf:   1.0000
INFO  [ecl/validation] sensor #1, prio: 100, state: OK
INFO  [ecl/validation]  val:   0.0020, lp:   0.0004 mean dev:  -0.0001 RMS:   0.0031 conf:   1.0000
INFO  [ecl/validation]  val:  -0.0020, lp:  -0.0033 mean dev:   0.0000 RMS:   0.0014 conf:   1.0000
INFO  [ecl/validation]  val:   0.0044, lp:   0.0041 mean dev:  -0.0000 RMS:   0.0011 conf:   1.0000
INFO  [sensors] accel status:
INFO  [ecl/validation] validator: best: 0, prev best: 0, failsafe: NO (0 events)
INFO  [ecl/validation] sensor #0, prio: 100, state: OK
INFO  [ecl/validation]  val:   0.0293, lp:   0.0279 mean dev:   0.0001 RMS:   0.0109 conf:   1.0000
INFO  [ecl/validation]  val:  -0.2623, lp:  -0.3164 mean dev:   0.0009 RMS:   0.0500 conf:   1.0000
INFO  [ecl/validation]  val:  -9.8288, lp:  -9.8239 mean dev:   0.0001 RMS:   0.0145 conf:   1.0000
INFO  [ecl/validation] sensor #1, prio: 100, state: OK
INFO  [ecl/validation]  val:   0.0361, lp:   0.0468 mean dev:  -0.0003 RMS:   0.0110 conf:   1.0000
INFO  [ecl/validation]  val:  -0.1933, lp:  -0.2559 mean dev:  -0.0002 RMS:   0.0477 conf:   1.0000
INFO  [ecl/validation]  val:  -9.7500, lp:  -9.7518 mean dev:   0.0003 RMS:   0.0172 conf:   1.0000
INFO  [sensors] mag status:
INFO  [ecl/validation] validator: best: 0, prev best: 0, failsafe: NO (0 events)
INFO  [ecl/validation] sensor #0, prio: 255, state: OK
INFO  [ecl/validation]  val:  -0.3583, lp:  -0.3579 mean dev:  -0.0000 RMS:   0.0032 conf:   1.0000
INFO  [ecl/validation]  val:  -0.4282, lp:  -0.4330 mean dev:  -0.0000 RMS:   0.0027 conf:   1.0000
INFO  [ecl/validation]  val:   0.4922, lp:   0.4932 mean dev:   0.0000 RMS:   0.0042 conf:   1.0000
INFO  [sensors] baro status:
INFO  [ecl/validation] validator: best: 0, prev best: 0, failsafe: NO (0 events)
INFO  [ecl/validation] sensor #0, prio: 75, state: OK
INFO  [ecl/validation]  val: 1014.4800, lp: 1014.5020 mean dev:   0.0070 RMS:   0.0593 conf:   1.0000
INFO  [ecl/validation]  val:  31.5700, lp:  31.5495 mean dev:   0.0225 RMS:   0.0265 conf:   1.0000
INFO  [ecl/validation]  val:   0.0000, lp:   0.0000 mean dev:   0.0000 RMS:   0.0000 conf:   1.0000
INFO  [sensors] Airspeed status:
INFO  [ecl/validation]  no data

INFO  [vehicle_acceleration] selected sensor: 3607306 (sensor_accel:0)
INFO  [vehicle_acceleration] offsets: -0.03656 -0.18259 0.10048
INFO  [vehicle_acceleration] scale: 0.99911 0.91098 1.00166
INFO  [vehicle_acceleration] bias: -0.00459 0.00511 -0.04384
vehicle_acceleration: cycle time: 30642 events, 519315us elapsed, 16.95us avg, min 14us max 396us 5.832us rms
vehicle_acceleration: interval: 30647 events, 4005.48us avg, min 1301us max 6783us 78.510us rms
vehicle_acceleration: sensor latency: 30651 events, 3809259us elapsed, 124.28us avg, min 115us max 2887us 26.389us rms

INFO  [vehicle_angular_velocity] selected sensor: 3672842 (sensor_gyro_control:1)
INFO  [vehicle_angular_velocity] offsets: 0.00266 0.02651 -0.02043
INFO  [vehicle_angular_velocity] scale: 1.00000 1.00000 1.00000
INFO  [vehicle_angular_velocity] bias: -0.00109 0.00118 0.00240
vehicle_angular_velocity: cycle time: 121605 events, 2317850us elapsed, 19.06us avg, min 17us max 49us 3.796us rms
vehicle_angular_velocity: interval: 121624 events, 1010.20us avg, min 677us max 1694us 352.201us rms
vehicle_angular_velocity: sensor latency: 121642 events, 2899439us elapsed, 23.84us avg, min 23us max 51us 2.222us rms

nsh> listener sensor_correction

TOPIC: sensor_correction
 sensor_correction_s
        timestamp: 139657561  (18.717096 seconds ago)
        gyro_offset_0: [0.0027, 0.0265, -0.0204]
        gyro_scale_0: [1.0000, 1.0000, 1.0000]
        gyro_offset_1: [0.0256, -0.0035, 0.0039]
        gyro_scale_1: [1.0000, 1.0000, 1.0000]
        gyro_offset_2: [0.0000, 0.0000, 0.0000]
        gyro_scale_2: [1.0000, 1.0000, 1.0000]
        accel_offset_0: [-0.0370, -0.1821, 0.1019]
        accel_scale_0: [0.9991, 0.9109, 1.0017]
        accel_offset_1: [0.0832, 0.1596, -0.0156]
        accel_scale_1: [0.9985, 1.0005, 0.9899]
        accel_offset_2: [0.0000, 0.0000, 0.0000]
        accel_scale_2: [1.0000, 1.0000, 1.0000]
        baro_offset_0: -2.2326
        baro_scale_0: 1.0000
        baro_offset_1: 0.0000
        baro_scale_1: 1.0000
        baro_offset_2: 0.0000
        baro_scale_2: 1.0000
        gyro_device_id: [3672842, 2360330, 0]
        accel_device_id: [3607306, 1442826, 0]
        baro_device_id: [3998482, 0, 0]
        gyro_mapping: [1, 0, 0]
        accel_mapping: [1, 0, 0]
        baro_mapping: [0, 0, 0]

dagar · 2019-09-30T02:11:48Z

I'd still like to split sensor_correction into several different simpler messages, but I'll save that for another pass.

dakejahl

Looks good!

dagar · 2019-09-30T17:43:42Z

TODO:

check that onboard calibration still works (what was handled in send_event)
skip starting temperature_compensation module entirely if not enabled (check parameters)

julianoes

I haven't fully grasped this yet. Need to look at it again with fresh eyes.

julianoes · 2019-10-16T16:48:11Z

src/modules/events/send_event.cpp

-	}
-
+	// TODO: do something with vehicle commands
+	// TODO: what is this modules purpose?


julianoes · 2019-10-16T16:53:49Z

src/modules/sensors/vehicle_acceleration/VehicleAcceleration.cpp

+
+			for (int correction_index = 0; correction_index < MAX_SENSOR_COUNT; correction_index++) {
+				if (corrections.accel_device_id[correction_index] == _selected_sensor_device_id) {
+					_sensor_correction_index = correction_index;


I'm not sure I understand this. Is _sensor_correction_index valid for the remainder? And if so, we only ever correct one of the sensors?

Is _sensor_correction_index valid for the remainder

No, unless there's a bug I am not seeing. SensorSelectionUpdate() is called everytime thru the loop in Run() and it will perform the check for data.device_id == selection.accel_device_id and if this fails it calls SensorCorrectionsUpdate which will update _sensor_correction_index. It's kinda tangled and messy and I don't particularly like it, but as far as I am aware this "works" (if it aint broke don't fix it?).

julianoes · 2019-10-16T16:56:28Z

src/modules/sensors/vehicle_angular_velocity/VehicleAngularVelocity.cpp

 			_offset = Vector3f{corrections.gyro_offset_2};
 			_scale = Vector3f{corrections.gyro_scale_2};

 		} else {
 			_offset = Vector3f{0.0f, 0.0f, 0.0f};
 			_scale = Vector3f{1.0f, 1.0f, 1.0f};
 		}
+	}
+
+	return false;


Why false?

I think this can be made to return void .. which means Start() could also return void seeing as how it's return value is unused.

BazookaJoe1900 · 2019-10-16T17:02:06Z

note that #13165 is related. and if what said on #13165 is true, it can be fixed with this pr

dakejahl

@dagar Have you tested this? (the onboard calibration routine).

dakejahl · 2019-10-24T17:39:22Z

src/modules/sensors/vehicle_acceleration/VehicleAcceleration.cpp

+
+			for (int correction_index = 0; correction_index < MAX_SENSOR_COUNT; correction_index++) {
+				if (corrections.accel_device_id[correction_index] == _selected_sensor_device_id) {
+					_sensor_correction_index = correction_index;


Is _sensor_correction_index valid for the remainder

No, unless there's a bug I am not seeing. SensorSelectionUpdate() is called everytime thru the loop in Run() and it will perform the check for data.device_id == selection.accel_device_id and if this fails it calls SensorCorrectionsUpdate which will update _sensor_correction_index. It's kinda tangled and messy and I don't particularly like it, but as far as I am aware this "works" (if it aint broke don't fix it?).

dakejahl · 2019-10-24T17:43:45Z

src/modules/sensors/vehicle_angular_velocity/VehicleAngularVelocity.cpp

 			_offset = Vector3f{corrections.gyro_offset_2};
 			_scale = Vector3f{corrections.gyro_scale_2};

 		} else {
 			_offset = Vector3f{0.0f, 0.0f, 0.0f};
 			_scale = Vector3f{1.0f, 1.0f, 1.0f};
 		}
+	}
+
+	return false;


I think this can be made to return void .. which means Start() could also return void seeing as how it's return value is unused.

dagar · 2019-11-24T21:06:46Z

Let's finish this.

dagar · 2019-11-24T22:12:07Z

TODO:

only run if enabled (TC_{A, B, G}_ENABLE or online calibrate
review vehicle_angular_velocity and vehicle_acceleration logic (recent lockstep bug)
test online calibrate

dakejahl · 2019-12-12T03:08:33Z

Tested successfully on the bench. Veriified online calibration (freezer + heat gun) as well as the continual updating of corrections as the sensors undergo temperature changes (heat gun).

We should remove TC_*_ENABLE parameters and use a single TC_ENABLE parameter instead. Right now the TemperatureCompensationModule always publishes corrections for all 3 sensors, however if a sensor does not have an associated calibration the offsets are 0 and the scales are 1, thus having no effect. Also, right now the subscribers of sensor_correction do not check the TC_*_ENABLE flags. Alternatively we could have the subscribers apply the updates on sensor_correction_updated as well as TC_*_ENABLE.

Thoughts?

dakejahl · 2019-12-12T03:12:52Z

The "published correction" was my debug printout to verify

dagar · 2020-01-21T02:10:46Z

Rebased on master and testing.

dagar · 2020-01-21T02:16:25Z

For systems not using temperature compensation this actually saves about 2 kB of memory (tested on fmu-v4).

nsh> free
             total       used       free    largest
Umem:       229568     178928      50640      47392  <- master
Umem:       229568     176736      52832      46288  <- PR

This also saves a small amount of cpu in the sensors module (maybe 0.2%).

dagar · 2020-01-21T02:32:10Z

I've verified sensor_correction on a few calibrated systems before and after, and this looks safe to merge.

Big thank you to @dakejahl for doing the heavy lifting here. This now makes the entire temperature compensation functionality modular and optional, so we can disable it where needed (eg fmu-v2) and save a fairly big chunk of flash and a little bit of memory. Having this out of the way will also help in refactoring the sensors module, which is one of the remaining pieces in the way of running multiple estimators.

julianoes · 2020-01-24T15:47:29Z

Nice @dagar!

dagar added Admin: Enhancement (improvement) 💡 Admin: Cleanup (housekeeping) 🧹 labels Aug 1, 2019

dagar added this to the Release v1.10.0 milestone Aug 1, 2019

dagar assigned dagar and dakejahl Aug 1, 2019

dagar force-pushed the pr-temperature_compensation_clean_up branch from 9b7a485 to 6bb7d2e Compare August 1, 2019 20:48

dagar commented Aug 1, 2019

View reviewed changes

weekly-digest bot mentioned this pull request Aug 11, 2019

Weekly Digest (4 August, 2019 - 11 August, 2019) #12682

Closed

weekly-digest bot mentioned this pull request Aug 18, 2019

Weekly Digest (11 August, 2019 - 18 August, 2019) #12733

Closed

dagar force-pushed the pr-temperature_compensation_clean_up branch from 13415c6 to 313d7f1 Compare September 15, 2019 15:35

weekly-digest bot mentioned this pull request Sep 22, 2019

Weekly Digest (15 September, 2019 - 22 September, 2019) #13002

Closed

dagar force-pushed the pr-temperature_compensation_clean_up branch 4 times, most recently from f4c3e35 to 7ed9c17 Compare September 30, 2019 02:06

dagar force-pushed the pr-temperature_compensation_clean_up branch 3 times, most recently from 3afb456 to 053ab20 Compare September 30, 2019 13:31

dakejahl reviewed Sep 30, 2019

View reviewed changes

weekly-digest bot mentioned this pull request Oct 6, 2019

Weekly Digest (29 September, 2019 - 6 October, 2019) #13096

Closed

This was referenced Oct 15, 2019

[ECL] Moving towards multi-sensor-per-same-measurement fusion #13127

Merged

Handle and log onboard_computer_status messages #13169

Merged

TSC21 requested review from dakejahl and julianoes October 15, 2019 14:37

TSC21 requested a review from bkueng October 15, 2019 14:37

dagar marked this pull request as ready for review October 16, 2019 16:46

dagar force-pushed the pr-temperature_compensation_clean_up branch from 053ab20 to ba3eed5 Compare October 16, 2019 16:57

julianoes reviewed Oct 16, 2019

View reviewed changes

weekly-digest bot mentioned this pull request Oct 20, 2019

Weekly Digest (13 October, 2019 - 20 October, 2019) #13236

Closed

dagar mentioned this pull request Oct 21, 2019

sensors split rc_update into new standalone module #13241

Merged

dagar force-pushed the pr-temperature_compensation_clean_up branch from ba3eed5 to 5c21cd5 Compare October 21, 2019 18:53

dakejahl reviewed Oct 24, 2019

View reviewed changes

dagar force-pushed the pr-temperature_compensation_clean_up branch from 5c21cd5 to 74cd307 Compare November 24, 2019 21:05

dagar force-pushed the pr-temperature_compensation_clean_up branch from 74cd307 to 3eb5de8 Compare November 24, 2019 22:05

dagar force-pushed the pr-temperature_compensation_clean_up branch from 3eb5de8 to b8bfd24 Compare November 25, 2019 15:10

dagar mentioned this pull request Dec 17, 2019

commander: move more static variables to class #13613

Merged

dagar force-pushed the pr-temperature_compensation_clean_up branch from e7e064b to 14958b7 Compare December 23, 2019 18:02

weekly-digest bot mentioned this pull request Dec 29, 2019

Weekly Digest (22 December, 2019 - 29 December, 2019) #13804

Closed

create temperature_compensation module

d0d1c0c

dagar force-pushed the pr-temperature_compensation_clean_up branch from bf7c223 to d0d1c0c Compare January 21, 2020 01:47

dagar changed the title ~~[WIP]: temperature_compensation module~~ temperature_compensation module Jan 21, 2020

dagar merged commit dc05cea into master Jan 21, 2020

dagar deleted the pr-temperature_compensation_clean_up branch January 21, 2020 02:42

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

temperature_compensation module #12601

temperature_compensation module #12601

dagar commented Aug 1, 2019 •

edited

Loading

dagar commented Aug 1, 2019 •

edited

Loading

dagar Aug 1, 2019

dagar Aug 1, 2019

dakejahl Aug 13, 2019 •

edited

Loading

dagar Aug 13, 2019

julianoes commented Aug 5, 2019

dagar commented Sep 30, 2019 •

edited

Loading

dagar commented Sep 30, 2019

dakejahl left a comment

dagar commented Sep 30, 2019 •

edited by dakejahl

Loading

julianoes left a comment

julianoes Oct 16, 2019

julianoes Oct 16, 2019

dakejahl Oct 24, 2019

julianoes Oct 16, 2019

dakejahl Oct 24, 2019

BazookaJoe1900 commented Oct 16, 2019

dakejahl left a comment

dakejahl Oct 24, 2019

dakejahl Oct 24, 2019

dagar commented Nov 24, 2019

dagar commented Nov 24, 2019

dakejahl commented Dec 12, 2019 •

edited

Loading

dakejahl commented Dec 12, 2019

dagar commented Jan 21, 2020

dagar commented Jan 21, 2020 •

edited

Loading

dagar commented Jan 21, 2020

julianoes commented Jan 24, 2020

temperature_compensation module #12601

temperature_compensation module #12601

Conversation

dagar commented Aug 1, 2019 • edited Loading

dagar commented Aug 1, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dakejahl Aug 13, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

julianoes commented Aug 5, 2019

dagar commented Sep 30, 2019 • edited Loading

dagar commented Sep 30, 2019

dakejahl left a comment

Choose a reason for hiding this comment

dagar commented Sep 30, 2019 • edited by dakejahl Loading

julianoes left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

BazookaJoe1900 commented Oct 16, 2019

dakejahl left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dagar commented Nov 24, 2019

dagar commented Nov 24, 2019

dakejahl commented Dec 12, 2019 • edited Loading

dakejahl commented Dec 12, 2019

dagar commented Jan 21, 2020

dagar commented Jan 21, 2020 • edited Loading

dagar commented Jan 21, 2020

julianoes commented Jan 24, 2020

dagar commented Aug 1, 2019 •

edited

Loading

dagar commented Aug 1, 2019 •

edited

Loading

dakejahl Aug 13, 2019 •

edited

Loading

dagar commented Sep 30, 2019 •

edited

Loading

dagar commented Sep 30, 2019 •

edited by dakejahl

Loading

dakejahl commented Dec 12, 2019 •

edited

Loading

dagar commented Jan 21, 2020 •

edited

Loading