RTL - and how difficult it is to return to stabilize

i have been using RTL a lot recently when i realise that my PID tuning has failed and manual control is difficult.  the thing is that im almost always flying in some wind and with a very small landing site surrounded by shrubs, rocks, or just plain uneven terrain.  as marco was suggesting here (http://www.diydrones.com/forum/topics/arducopter-2-3-released?commentId=705844%3AComment%3A776259), a completely auto RTL and landing can be risky in these conditions and he suggests an alternative method performing auto approach but manual landing.

 

basically i agree completely with this, im much happier controlling the last metre or two of descent in stabilize giving me the option to choose exactly where i land, and more importantly killing throttle the moment it touches down.  

 

in theory i think this could be accomplished without any more complication in the process of auto land - for me the only thing that is needed is for arducopter to return to stabilize the moment RTL is disengaged.  currently for some reason this is very difficult, i often try to disengage RTL at 2 metres above ground but normally it seems to completely ignore the mode change and continues with its auto landing.  several times it has tipped over on landing and even being in stabilize and cutting the throttle completely it continues to spin props furiously trying to dig a hole in the ground.

 

i will look through the code to see if i can understand it, but what are the conditions at this time to exit RTL?  i really think that just facilitating the exit from this mode we accomplish everything marco suggested.

 

--------------------

 

and now onto the second part of my post- its entirely my fault for not checking the props carefully before flying, but ultimately is the result of a failed auto landing.

 

obviously in the previous "tipped over, motors thrashing into the ground" auto landing a prop was loosened, and 10 minutes later came off in flight.  as i have been tuning pid values and in many cases failing completely, when i saw it suddenly oscillate and flip over onto its side i thought it was simply a failure of my tuning and really thought there was nothing i could do to regain control.  as it rolled over, glinted briefly in the last rays of the evening light, and dropped like a stone out of view down the mountainside i simply cut the throttle and sent "disarm motors" while listening for the crash.

 

it was only as it shot out of view that i noticed the prop fluttering in the air in the original flight path - if i had given it some throttle and tried to counter the roll i may even have been able to fly it (hexa), or at least get it closer and perform a controlled crash landing.  as it was it fell un-powered 50 or 60 metres down towards the barren, rocky slope.

 

we started looking for it and after about twenty minutes and some fairly scary climbing located the starboard nav lights poking out of a plant.  it was located but unreachable from above, we photographed the site as darkness fell and arranged to meet the next morning to attempt the recovery from below:

3690914656?profile=original

 

so the hexa had to spend the night on the mountainside, with APM and nav lights powered up, it even rained.  after a very close ispection of the site with google earth and a lot of worrying about the possibility of the LiPo catching fire we arrived the next morning with a 70m climb before us:

3690914595?profile=original

 

we were lucky and found the spot pretty much straight away, amazingly, with bare rock all around it had come down bang in the centre of the plant (this is an invasive species displacing native flora, so we didnt feel bad about trimming it to get the hexa out):

3690914605?profile=original

3690914678?profile=original

 

damage is limited to two arms and two broken props - far beyond my expectations!

 

APM has powered up and shows the horizon correctly, all motors spin up fine and react to maintain level.  even the LiPo looks like it might survive, cells were at 2.7, 2.7, and 2.9 - it has charged and balanced ok, but i will watch this very closely when it flys again.  just need to replace those arms, dozens of nylon screws, and square everything up again.

 

all in all a very lucky escape - i thought we would be picking up pieces over a wide area :D

 

james

 

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –

Replies

  • Developer
    James, I am very sorry for what has happened, unfortunately although part of the game but see his quad precipitate is never good, and tells you who has a good experience with this, always due bugs... :-/
    One thing I noticed in your assembly is that you used that kind of "arches" to attach the propeller to the engine, thing I personally hate it because if you put the good threadlocker is unscrewed very easily.
    I put them as security, but beneath them is the locking nut to secure the propeller, absolutely.
    What time did I ask you to retrieve the logs from your APM and post it here, we want to check the problem with RTL and "exit from autolanding".
    With the simulator i'm always able to inhibit autolanding during the descent phase, so it's very strange that you will not be successful, and the log is easy to see what happened.
    Let's say you were also lucky in bad luck, the weeds have saved your quad, I do not want to imagine what would have happened if he had fallen on one of those houses...
    In the new version of ArduCopter 2.3.1 there is the possibility to inhibit the automatic landing, what I personally regard as almost useless for all the quad quite "important".
    Do not get me wrong, I mean that a small quad manages to land safely, a large can easily tip over the ground.
    For this reason I suggested the good alternative of "auto-approach", which for now has not been taken into account, but I'll try to implement and testing it and provide the code for those who want to try.
    Personally I will leave active automatic landing only at failsafe, this is a new thing in 2.3.1.

    We await your log.

    Bests

    Marco
This reply was deleted.

Activity