Holybro UART optical flow (thoneflow driver) severe oscillations position mode [Bug]

[dirksavage88]

Describe the bug

I've tested the holybro uart sensor on a 210mm wheelbase quad and found severe position oscillations when using 1.14, main, and even going back to 1.13.3 stable. I swapped the holybro sensor for the Ark can flow sensor with updated cannode firmware (recent main) that includes the gyro bias estimation fixes. The ark flow worked great.

See this forum discussion:
https://discuss.px4.io/t/holybro-optical-flow-sensor-thoneflow-oscillations-on-main/34768

To Reproduce

Enable the thoneflow driver, set the OF, EKF2, and rangefinder params, and take off in position mode

Expected behavior

Optical flow position mode holds stable position with minimal oscillation

Screenshot / Media

Here is a pitch angle plot from the holybro sensor:

Same exact hardware setup except the ark flow:

Flight Log

Holybro (like a merry go round) : https://review.px4.io/plot_app?log=94031e37-9350-4b03-a76e-19fdd9cdcf4d
Ark can flow (I'm not moving): https://review.px4.io/plot_app?log=7ea5d915-144a-45ff-971c-76aa3d33732e

Software Version

tested with similar results on stable, 1.14 beta, recent main

Flight controller

holybro 6c mini

Vehicle type

None

How are the different components wired up (including port information)

No response

Additional context

No response

[DronecodeBot]

This issue has been mentioned on Discussion Forum for PX4, Pixhawk, QGroundControl, MAVSDK, MAVLink. There might be relevant details there:

https://discuss.px4.io/t/holybro-optical-flow-sensor-thoneflow-oscillations-on-main/34768/1

[dagar]

[dirksavage88]

As requested @dagar @bresch here are three logs that show an attempt at a box pattern as well as spinning by hand in position mode while armed (but landed):

1. box pattern: https://review.px4.io/plot_app?log=08d11193-1a40-498f-b366-cd56d09862eb
2. Spinning 1: https://review.px4.io/plot_app?log=9ed53bbc-df6c-41e1-95c2-882b4cd0bfda
3. Spinning 2: https://review.px4.io/plot_app?log=6ad9c489-5c97-4fcc-b172-a645d395fc93
4. EKF2 replay from box pattern: https://review.px4.io/plot_app?log=27ad9e20-8aa6-4d33-a137-3d20e8fe3348

[jfaure6]

Hello @dirksavage88 !

I observe exactly the same phenomenon with the same hardware (pmw3901, thoneflow driver in UART).
My setup is Holybro's px4vision in v1.5. Note that I also have the Holybro PX4Vision in v1, with the PMW3901 in SPI, but with this configuration, I do not observe any oscillations.
You can view the logs:
v1.5 : https://review.px4.io/plot_app?log=d2530b83-3116-4b63-a292-8a3d440a3993
v1 : https://review.px4.io/plot_app?log=e8908b3a-7745-41a9-859f-4bbdddd4349a

[bresch]

Looks like the scale in the driver might be wrong. Try divide those scales by 2 (replace 3.52e-3 by 1.76e-3):

PX4-Autopilot/src/drivers/optical_flow/thoneflow/thoneflow_parser.cpp

Lines 136 to 137 in 8aab3e8

	flow->pixel_flow[0] = static_cast<float>(delta_x) * (3.52e-3f);
	flow->pixel_flow[1] = static_cast<float>(delta_y) * (3.52e-3f);

[mmalecki]

This patch has greatly improved OF performance with this sensor for me:

Before, note the oscillations with no RC input:

After, lack of earlier oscillations:

Thanks @bresch!