Pr wind est work queue #9079

RomanBapst · 2018-03-14T14:14:05Z

This is a replacement of #9041 and implements the standalone wind and airspeed scale factor estimator as a work queue. Previously, the estimator was living in the sensors module which did not really make sense.

Advantages:

the estimator is now a separate module and can be run only for VTOL and fixed wing configs
we can extend this module to take data of multiple airspeed sensors and run multiple estimators and then use this information for airspeed failure detection

Disadvantages:

we have some overhead in terms of memory due to the work queue and the fact that we need to subscribe to three topics, one of them is local_position which is rather large

@dagar @priseborough FYI

bkueng · 2018-03-14T14:22:07Z

src/modules/wind_estimator/wind_estimator_main.cpp

+	/* schedule a cycle to start things */
+	work_queue(LPWORK, &_work, (worker_t)&WindEstimatorModule::cycle_trampoline, nullptr, 0);
+
+	// wait until task is up & running (the mode_* commands depend on it)


(the mode_* commands depend on it) can be removed

bkueng · 2018-03-14T14:22:25Z

src/modules/wind_estimator/wind_estimator_main.cpp

+	static int print_usage(const char *reason = nullptr);
+
+	/**
+	 * run the main loop: if running as task, continuously iterate, otherwise execute only one single cycle


This comment is incorrect as well

dagar · 2018-03-14T15:13:58Z

Disadvantages:

we have some overhead in terms of memory due to the work queue and the fact that we need to subscribe to three topics, one of them is local_position which is rather large

The work queue should be a slight improvement. Previously in sensors you were keeping copies of local_position, attitude, airspeed in the sensors object, and updating the subscriptions constantly (sensors task usually running at 250 Hz). Now you only pay for it during each iteration of the wind estimator at a significantly lower rate.

RomanBapst · 2018-03-15T08:36:26Z

@PX4/testflights Could you please give this a test flight, preferably when it's windy.
Thanks!

RomanBapst · 2018-03-15T08:37:36Z

@PX4/testflights I forgot to mention that you can't test it on fmu-v2 due to lacking flash space.

r0gelion · 2018-03-15T14:56:01Z

@RomanBapst our FW has a fmu-v3 (Pixhawk mini). It's a rainy day, but as soon as the weather allows it we will test it.

Regards.

dagar · 2018-03-16T21:02:36Z

Merge the ECL side and update this PR?

r0gelion · 2018-03-20T16:02:13Z

@RomanBapst @dagar Is this ready for flight testing?

RomanBapst · 2018-03-20T16:47:20Z

@r0gelion It's good that you ask, no this is not ready. I just came back from a test flight and there seems to be a problem with the work queue which i need to discuss with @dagar
I'll ping you as soon as we solve it.

dagar · 2018-03-21T04:30:51Z

The issue here was mixing different work queues for the initial and repeated cycles. See #9122

RomanBapst · 2018-03-21T10:35:00Z

@dagar Thanks! I just tested it and it works fine now. Below are the results of the wind estimate using only sideslip fusion (no airspeed data).

These were the forecasted wind conditions at the time of the test flight:
wind: NE direction @ 10 knots (5.1 m/s)

Luckily there was a wind sock near the flying field which confirmed the wind coming out of NE.
The big building to the left of the wind sock lies roughly in north direction.

Results of the estimator when flying circles:

Wind component estimate and absolute value:

Sideslip innovation & innovation standard deviation

Ground speed and vehicle heading

The absolute value estimate of 5 m/s seems quite accurate and also the wind components suggest wind out of Northeast, North-northeast.
Two separate test flight were conducted and the results of both were similar:

https://logs.px4.io/plot_app?log=a842cef4-03f8-4722-92db-6bf38f7d7788
https://logs.px4.io/plot_app?log=46e98042-87c9-4117-9f8f-fe4536ff7ca6

- true airspeed innovation and variance - sideslip angle innovation and variance Signed-off-by: Roman <[email protected]>

work queue Signed-off-by: Roman <[email protected]>

Signed-off-by: Roman <[email protected]>

…ance Signed-off-by: Roman <[email protected]>

…lash space Signed-off-by: Roman <[email protected]>

Signed-off-by: Roman <[email protected]>

dagar · 2018-03-21T13:43:00Z

For reference here's what it looks like running in the low priority work queue (first log).

Which means that the cycle is being called every 101317us on average (~ 10Hz), and average runtime for each cycle is 159 microseconds. The rate skews a bit because the end of every cycle after the filter update schedules another 100 ms later.

dagar · 2018-03-21T17:08:26Z

Update ecl to current master and this should be good to go.

… tests Signed-off-by: Roman <[email protected]>

- if it's not estimating wind it won't publish it below (was fixed) Signed-off-by: Roman <[email protected]>

- need to make some flash space for fmuv2 to avoid user confusion about why it wind estimator only runs on some platforms Signed-off-by: Roman <[email protected]>

RomanBapst mentioned this pull request Mar 14, 2018

Pr wind est rebased #9041

Closed

bkueng reviewed Mar 14, 2018

View reviewed changes

RomanBapst force-pushed the pr-wind-est-work-queue branch from 52f3cc2 to aecf5c6 Compare March 14, 2018 14:45

RomanBapst force-pushed the pr-wind-est-work-queue branch from 85c9c33 to a6ca4bd Compare March 15, 2018 08:43

RomanBapst and others added 13 commits March 21, 2018 12:12

wind estimate msg: added more relevant data

efe4355

- true airspeed innovation and variance - sideslip angle innovation and variance Signed-off-by: Roman <[email protected]>

added module which implements wind & airspeed scale factor estimator via

ba4bc83

work queue Signed-off-by: Roman <[email protected]>

cmake configs: added wind estimator to relevant configs

10140ac

Signed-off-by: Roman <[email protected]>

start wind estimator for relevant platforms

ac4fcd4

Signed-off-by: Roman <[email protected]>

wind_estimator: unsubscribe from topics and unadvertise wind topic

c3be903

Signed-off-by: Roman <[email protected]>

wind_estimator: set missing task_id

c57f8ba

Signed-off-by: Roman <[email protected]>

wind_estimator: cleanup

4747f6e

Signed-off-by: Roman <[email protected]>

ekf2: advertise wind topic early to make sure ekf2 get the first inst…

e53b3c5

…ance Signed-off-by: Roman <[email protected]>

cmake configs: don't build wind estimator for fmu-v2 due to lack of f…

66be4a1

…lash space Signed-off-by: Roman <[email protected]>

wind_estimator: use ModuleParams class to handle parameters

a45f337

Signed-off-by: Roman <[email protected]>

wind estimator: fixed minor comments

ce5ec53

Signed-off-by: Roman <[email protected]>

wind_estimator fix LPWORK/HPWORK confusion and cleanup

c4c0aa6

wind_estimator trim module documentation to fit in limit

3521b06

RomanBapst force-pushed the pr-wind-est-work-queue branch from c4d460a to baf1ffa Compare March 21, 2018 11:20

RomanBapst force-pushed the pr-wind-est-work-queue branch from baf1ffa to 989934d Compare March 21, 2018 17:04

RomanBapst added 4 commits March 21, 2018 18:31

updated ecl for airdata module

250872e

wind estimator: added airspeed and sideslip gate sizes for innovation…

a40ceee

… tests Signed-off-by: Roman <[email protected]>

ekf2: remove early advertising of wind estimate topic

2614cbe

- if it's not estimating wind it won't publish it below (was fixed) Signed-off-by: Roman <[email protected]>

fw & vtol startup scripts: don't start wind estimator for now

697bfc3

- need to make some flash space for fmuv2 to avoid user confusion about why it wind estimator only runs on some platforms Signed-off-by: Roman <[email protected]>

RomanBapst force-pushed the pr-wind-est-work-queue branch from 009cde4 to 697bfc3 Compare March 21, 2018 17:32

RomanBapst merged commit e504dc8 into master Mar 21, 2018

RomanBapst deleted the pr-wind-est-work-queue branch March 21, 2018 17:47

mrpollo mentioned this pull request Jun 5, 2018

v1.8 Changelog scratchpad #9593

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Pr wind est work queue #9079

Pr wind est work queue #9079

RomanBapst commented Mar 14, 2018 •

edited by AuterionWrikeBot

Loading

bkueng Mar 14, 2018

bkueng Mar 14, 2018

dagar commented Mar 14, 2018

RomanBapst commented Mar 15, 2018

RomanBapst commented Mar 15, 2018

r0gelion commented Mar 15, 2018

dagar commented Mar 16, 2018

r0gelion commented Mar 20, 2018

RomanBapst commented Mar 20, 2018

dagar commented Mar 21, 2018

RomanBapst commented Mar 21, 2018

dagar commented Mar 21, 2018

dagar commented Mar 21, 2018

Pr wind est work queue #9079

Pr wind est work queue #9079

Conversation

RomanBapst commented Mar 14, 2018 • edited by AuterionWrikeBot Loading

bkueng Mar 14, 2018

Choose a reason for hiding this comment

bkueng Mar 14, 2018

Choose a reason for hiding this comment

dagar commented Mar 14, 2018

RomanBapst commented Mar 15, 2018

RomanBapst commented Mar 15, 2018

r0gelion commented Mar 15, 2018

dagar commented Mar 16, 2018

r0gelion commented Mar 20, 2018

RomanBapst commented Mar 20, 2018

dagar commented Mar 21, 2018

RomanBapst commented Mar 21, 2018

dagar commented Mar 21, 2018

dagar commented Mar 21, 2018

RomanBapst commented Mar 14, 2018 •

edited by AuterionWrikeBot

Loading