I think that one very important point we are missing is that the optical flow sensor needs to installed with VIBRATION insulation just like the IMU. I have the original 3DR optical flow sensor and Lidar Lite V2. When I test the senor just by waving the multicopter by hand the Of.flowX/OF.bodyX/IMU.GyrX perfectly match. However when inflight the OF.flowX is way off the other two values which match.
Log when waved by hand.
Log when flying.
I have been thinking hard what could be the reason. The only difference I can think of is vibration when the copter is flying. I have just attached the optical flow to the base of my copter with some hot glue and attached a zip tie. The zip tie is to prevent the sensor from falling, just in case the hot glue failed.
Tomorrow I will attach the sensor using some nice vibration isolation foam and try again. Do you guys think that this will help?
I did the camera focus with a checkerboard and text printed on A4 paper on the ground. So I doubt this it it.
I was flying above mats with yellow colour and colourful spots in them. Possible the contrast is not good enough. I'll check with the high def video stream. I could try printing several A0 size checkerboards and pinning them to the floor.
We have TL lights in the building. Perhaps strobing is possible. That would explain why both files are noisy as I did the calibration above a checkerboard floorpanel. I will have to test outdoor to be sure, but the weather is too bad the next few days.
I'll also add more damping to the optical flow mount.
I've tested the optical flow setup on my DJI F550 Hexa with an i2c external sonar and PX4FLOW module with Copter V3.4-dev (89c14718) which I downloaded today.
I've calibrated the optical flow manually according to the following wiki page and it looks ok.
The test flight log link with EK2 and optical flow is below. It seems the the EK2 and optical flow are doing well, except that during flight the hexa starts drifting in various directions and, as someone posted before, it reacts sluggishly to pilot inputs.
I suspect that the loiter PID is not tuned well enough as NTUN's DesVelX vs VelX and DesVelY vs VelY are not matching. But before I try playing with PID parameters, does anyone have another idea why this may be? Can I be sure that the optical flow and EK2 are indeed working as I suspect? The loiter wiki page says its very unlikely that the loiter PID has to be tuned differently from default values. Therefore I'm not sure.
I'm not clear with wiki on how to calibrate OF. It just says "calibrate" but not more explanation. Could you help me to get further information on calibration?
Replies
Yes, what I did
GPS=disconnected
GPS.Mag=connected
FlightMode=Loiter
and it was perfectly armed.
In case it helps, last week I added a bit more information re the pre-arm checks to the px4flow wiki page.
Yeh, just downloaded via MP
I think that one very important point we are missing is that the optical flow sensor needs to installed with VIBRATION insulation just like the IMU. I have the original 3DR optical flow sensor and Lidar Lite V2. When I test the senor just by waving the multicopter by hand the Of.flowX/OF.bodyX/IMU.GyrX perfectly match. However when inflight the OF.flowX is way off the other two values which match.
Log when waved by hand.
Log when flying.
I have been thinking hard what could be the reason. The only difference I can think of is vibration when the copter is flying. I have just attached the optical flow to the base of my copter with some hot glue and attached a zip tie. The zip tie is to prevent the sensor from falling, just in case the hot glue failed.
Tomorrow I will attach the sensor using some nice vibration isolation foam and try again. Do you guys think that this will help?
Hi Paul,
Thanks for the fast reply.
I did the camera focus with a checkerboard and text printed on A4 paper on the ground. So I doubt this it it.
I was flying above mats with yellow colour and colourful spots in them. Possible the contrast is not good enough. I'll check with the high def video stream. I could try printing several A0 size checkerboards and pinning them to the floor.
We have TL lights in the building. Perhaps strobing is possible. That would explain why both files are noisy as I did the calibration above a checkerboard floorpanel. I will have to test outdoor to be sure, but the weather is too bad the next few days.
I'll also add more damping to the optical flow mount.
Thanks again for the input!
Jon
Hi,
I've tested the optical flow setup on my DJI F550 Hexa with an i2c external sonar and PX4FLOW module with Copter V3.4-dev (89c14718) which I downloaded today.
I've calibrated the optical flow manually according to the following wiki page and it looks ok.
http://copter.ardupilot.com/wiki/common-px4flow-overview/
Log download link: https://www.dropbox.com/s/j3rrwqfokm4qbo8/Optical%20flow%20manual%2...
The test flight log link with EK2 and optical flow is below. It seems the the EK2 and optical flow are doing well, except that during flight the hexa starts drifting in various directions and, as someone posted before, it reacts sluggishly to pilot inputs.
Log download link: https://www.dropbox.com/s/r1y79y8ty31ix9f/EK2%20and%20optical%20flo...
I suspect that the loiter PID is not tuned well enough as NTUN's DesVelX vs VelX and DesVelY vs VelY are not matching. But before I try playing with PID parameters, does anyone have another idea why this may be? Can I be sure that the optical flow and EK2 are indeed working as I suspect? The loiter wiki page says its very unlikely that the loiter PID has to be tuned differently from default values. Therefore I'm not sure.
http://copter.ardupilot.com/wiki/loiter-mode/
Cheers,
Jon
P.S. My apologies as the attachment limit of 7MB was too low to attach the logs directly to this post.
Hi,
I would like to implement Optical Flow into ArduRover, emplying laser mouse and modified lens to get sharp image underneath.
Distance is constant so no need to install distance sensor (sonar, Lidar lite)
PC laser mouse is highly sophisticated Optical Flow machinery and algorithm.
Pls let me know your opinion and your experience with laser mouse employed as Optical Flow controler.
I also have this problem. What works is logging in via NSH and rebooting the px4flow driver:
-Use mission planner
-Enable advanced mode, then click the Terminal tab
-Select NSH
-Connect
-Type "px4flow start"
-px4flow status (or was it info) should tell you the driver is running
Hi Paul,
I'm not clear with wiki on how to calibrate OF. It just says "calibrate" but not more explanation. Could you help me to get further information on calibration?
Thank you Paul, as usual, for useful information!
-
4
-
5
-
6
-
7
-
8
of 11 Next