Here some videos samples of 8 full AUTO-LANDING tests flights that I have done with the firwmare ArduCopter v2.4xp1 (JLN mod). I have updated the official ArduCopter v2.4 firmware with a special Auto-Landing mod that I have developped and tested successfully in the previous v2.2xp2 release.
In the video you will see, full tests of the:
- STABILIZE mode (full manual flight evolutions),
- LOITER mode (GPS position hold),
- and 4 samples of FULL AUTO-LANDING with the v2.4xp1
The quadcopter is firstly set in LOITER mode and then the AUTO-LANDING is trigged by the Switch 7 of the transmitter. Setup: ArduPilot Mega v2 + Firmware ArduCopter v2.4xp1, Frame: Quadcopter DJI F450 Flame Wheel (X mode)
Here some videos samples of full AUTO-LANDING tests flights in HIL mode on AeroSIM-RC 3.83 with a Raptor 30 type helicopter (H1) done with the firwmare ArduCopter v2.4xp2 (JLN mod). The official v2.4 firmware has been updated with a special Auto-Landing mod which uses only the BARO sensor. Tested with an ArduPilot Mega v1.4 connected in HIL mode on AeroSIM-RC v3.83 through the APM Mission planner.
more infos at: http://diydrones.com/profile/JeanLouisNaudin
Comments
YES, this is a linear regression of course, this is commonly used in Machine Learning process... This is that I have used here. I have recorded a lot of manual landing done by mylsef with my quad and then computed the logged data to find the parameters P1 and P2 with the MLA... This is an easy and secure method ;-)
Excellent result Jean Louis!
Now i try in the sim...
A high P1 increase dramatically the descent rate (squared), a higher P2 gives a smooth descent rate.
I recommend you to tests on grass at above 2 or 3 meters, this is useful for tuning safety the P1, P2 parameters Vs the weight of your quad.
Regards, Jean-Louis
:) I'll be making some tests tomorrow, I only have 1 battery now, since the other damaged on a crash :| grrrr
Jean-Louis, since I don't know how the computation is made, and how the algorithm works, and what x and y are, if I have a 1600 or 1500gr quad, how do changing the P1 and P2 params affect the drop ? Meaning, higher P1 will give me slower or faster descent ? and higher P2 ? I'm sorry for bugging you but I don't know how the calculus are made.
@Rui: I have forgotten to say how to use the SW on CH7 to activate the AUTO_LAND.
In the source it is set with:
#define CH7_OPTION CH7_LAND
or in the MP by the command CH7_OPT = 9
In flight, if you want to do a simple test (no RTL is required for the test), set the quad in LOITER mode at above 2 or 3 meters above the ground, and then set the CH7 to ON then OFF, the AUTO_LAND is activated... Then wait and see (and cross the fingers...). Don't forget to switch back to MANUAL (stabilize).
Good luck...
@ARHEXA: it use the BARO... this is the challenge here... With sonar is too much easy ;-)
Regards, Jean-Louis
does it use the sonar to auto land ?
it would be really cool to use the sonar:
drop fast while above 4 meters, then slow down the descend,
and do the last meter or 50cm really slow and gentle
Here the details about the AUTO_LAND curve computed by the ML algorithm used during the learning phase that I have done for the landing with my quadcopter. ( weight = 988 g). You see the computation of the parameters P1 and P2.
Regards,
Jean-Louis
Very cool Jean-Louis. Thanks for sharing.
Hello Rui,
As you may notice there are 3 parameters in the yellow rectangle (above):
AUTO_LAND_P1 = 175
AUTO_LAND_P2 = 72
CH7_OPT = 9
AUTO_LAND_P1 and AUTO_LAND_P2 are parameters used by my ML Alogrithm for the final touchdown.
These two parameters are linked to the weight on the Power/weight ratio of the multicopter. These 2 parameters can be tuned for heavier or lighter copters so as to get a kiss landing.
The CH7_OPT = 9 is for the Switch on CH7 and the 9 is for activating the AUTO-LAND.
FYI: This AUTO-LAND feature is also used by the RTL.
Regards, Jean-Louis
-
1
-
2
-
3
-
4
of 4 Next