Acro Trainer Mode and Frame Reccomendation

[Brian DeBusk]

I'm considering trying to learn flying in Acro mode and I'm also intrigued by the Acro Trainer mode.  It sounds like the perfect way to learn without crashing a lot of hardware.

1) How mature is the Acro trainer feature?  Is this something that has been well tested and debugged at some point?

2) And how stable is ArduCopter's Acro mode in general?  Is that fairly debugged and tested?

3) Given all the recent changes with the EKF, would any of that impact the Acro and Acro Trainer features?  Would I be more prone to DCM/EKF errors and triggering forced landings?  Should I completely turn the EKF off, or does Acro mode do that automatically?

and finally...

4) What frame would your recommend for Acro training?  Obviously a larger and/or more expensive frame makes a crash more expensive.  I was thinking about something small, but not so small that I couldn't mount a Pixhawk on it well.  What do you think about a QAV 400 or a Discovery TBS Pro?  I seriously doubt a QAV 250 could hold a Pixhawk, or I'd even consider that.

Thanks in advance,

Brian

 

[Leonard Hall]

Hi Brian,

Acro and trainer is more mature than position hold and just a reliable as Stabilize.

As for Frame, I would suggest something with 8 to 10" props. I have a QAV250 with the full pixhawk and fpv kit on it but it is a tight, tricky build.

As you are practising your acro, make sure you have Stabilize and RTL easily accessible on your TX. Your escape response should be stabilize and full throttle, in that order.

Start with full angle limiting and your roll and pitch levelling set to something like 0.5. Then slowly reduce it down to 0.1 or 0. Then when you are comfortable you can turn off the angle limiting. You can also switch that on and ch7 or ch8 switch.

Good luck.

Leonard

[Robert Lefebvre]

On point 2), I've done extensive flying in non-Trainer acro with my helicopters, which is exactly the same program as for multirotors.  It works very well, and there are no bugs.  There were a few during development, but it was fixed.  It works very very well.

Point 3) is more complicated to answer.  I fly Acro with EKF.  My opinion, is that the EKF is actually better than the DCM.  However, EKF will also quit very early, and fall-back to DCM automatically.  And it doesn't take much.  The problem is the EKF is actually trying "too hard" to get a perfect heading estimate, and gives up early.  When it fails-back to DCM, the DCM is usually worse.  

There are two things that cause this.  First, the GPS signal can be lost when upside-down. I'm not sure if this is a big problem for multi-rotors, but since I can fly the heli upside-down, it happens.  But also, constant flips and rolls and rotations, causes it to fail because the algorithm does not take into account for scale-factor errors of the sensors.  So if you spin it for too long, an error builds up, and the EKF just quits.  The DCM has the same error, but it does not care too much.

Acro on Arducopter is really good for learning.  You just can't push it too hard.  Doing one flip, then resting for a couple seconds for flipping again, is all that's needed.  Where I get into trouble is doing 5+ successive flips, just keep going around and around.  That's where EKF fails-back to DCM.

High speeds, banking and yanking are completely not a problem.  It's just continuous flipping, or being upside-down that causes the problems.

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Josh Welsh]

Just to add to Leonard and Rob’s comments, Acro is very solid for me as well, and is the flight mode of choice when I am flying even my gimbal ships in fpv.  Also, note that Rob said don’t push it too hard – which means you can push it, but, at some point when you start getting very aggressive with flips rolls in succession, that’s when you may have problems.  I’ll regularly pop up high, point the nose down via 270 backflip and do 3-4 rolls with an ACRO_RP_P value of 11 or 12 coming out of the maneuver no problem.  The mode is mature for sure.

 

Leonard is spot on with the qav250, his is a bit tight and a bit heavy but it still flies fine.  I’ve seen a few others that are fully decked out and they are certainly crowded – but they fly just fine.  I’ve run the pixhawk on the qav400 and it is fine but can be tricky to autotune due to the flex inherent to the dirty frame.  The frame I use to test out Acro for Leanord is the Cinetank Mk.2 from flyingcinema.com and it is a very solid design.  I’ll recommend that frame all day long.  You do have to mind the arms from time to time after a crash to make sure they don’t twist the motor’s vertical axis even slightly (this will throw off any flight controller not just the pixhawk) but overall it is a great frame, and is the frame I used for the demo videos on the acro wiki page.  I can’t speak to the TBS Disco but I know those arms are also a little flexible so see potential for not-ideal autotune, but I’m sure you can check the forums to see what people with that frame are saying about those. 

 

Good luck on whatever you decide, and on your build!

[Robert Lefebvre]

Acro RP P of 11 or 12?  Nice!  I've been using 10, wasn't sure if it could go higher.  Actually, I'm using I think 6 or 8 as standard on my heli now.  Having negative pitch means you don't need to flip as fast. ;)

[Josh Welsh]

J

 

I thought the cap was at 10 as well, but in talking with Leonard I’ve found it can go quite a bit higher.  I just wish there was an easy way for me to see what I’m actually achieving in rate deg/sec but doing that is above my head at the moment.

 

Yeah I started flying heli very recently (TREX 550E stretched to 600 with BeastX) and I’ve noticed the roll and pitch rates are quite a bit slower than what I’m used to, but exactly like you said you don’t necessarily want it as fast when you can go negative.

 

I’ve considered getting another 500-ished size heli and working with you on setting it up with a pixhawk, but I’ve gotta calm down on the number of toys in the garage J

 

From: drones-...@googlegroups.com [mailto:drones-...@googlegroups.com] On Behalf Of Robert Lefebvre
Sent: Monday, January 12, 2015 11:50 AM
To: drones-discuss
Subject: Re: [drones-discuss] Re: Acro Trainer Mode and Frame Reccomendation

 

Acro RP P of 11 or 12?  Nice!  I've been using 10, wasn't sure if it could go higher.  Actually, I'm using I think 6 or 8 as standard on my heli now.  Having negative pitch means you don't need to flip as fast. ;)

 

[Brian DeBusk]

All great input -- very much appreciated.

I remember in the early 3.2 betas, bad enough EKF errors would trigger an auto-land.

Would that still happen in Acro and Acro Trainer mode?  Let's say I keep the EKF turned on and do single flips, but nothing crazy (i.e. 5+ flips in a row). 

If the EKF does throw an error, will it try to auto-land or is that just in Loiter mode?

Thanks in advance,

Brian
 

[Josh Welsh]

Would that still happen in Acro and Acro Trainer mode?  Let's say I keep the EKF turned on and do single flips, but nothing crazy (i.e. 5+ flips in a row). 

Single flips will not produce any problems, so feel free to flip/roll away!  EKF is safer than DCM, especially now.

If the EKF does throw an error, will it try to auto-land or is that just in Loiter mode?

 

It'll fail to DCM transparently and there are safe guards to try to keep it from going nuts, but since you're flying rate-based control rather than autopilot-assisted attitude control, you should be generally ok. 

Honestly, I think you're possibly overthinking/worrying too much about the EKF failing.  I hope that doesn't sound too forward, but I just wanted to stress that his isn't something that is a regular occurrence.  It happens for Rob more because I think he's pushing it trying to find where it breaks and he's doing it with TradHeli with a lot of maneuvering and high g forces - that's quite a ways further down the road than you trying your first flip and roll.  I encourage you to go have fun and I'd be very surprised to hear you had any issues with it, assuming everything is set up properly.

Good luck, I'm looking forward to hearing how you do after you try!

[Robert Lefebvre]

No, it won't autoland... unless something changed that I'm not aware of.  Last time I flew it was maybe September.

It can lead to a 10° jerk in Acro, it's complicated, but not a huge problem.

Actually Josh, it's completely fine with high-G forces.  I've pulled 6 Gs pullups, and 2-3G loops, no problem with that at all.  It's the rotation rate, continuous rotation rate, that causes the problem.  Just don't do more than 3 consecutive flips, and you should be fine.  You really can't do that on a multi anyway.

[Josh Welsh]

Ah, ok thanks Rob, I remember a thread a while back where I thought the 6g’s were contributing.  Thanks for clearing me up on that.

 

FYI, You can definitely do more than 3 consecutive flips on a multrotor.  Yes you’ll lose altitude, but, its possible.  Turn your ACRO_RP_P up past 10, man! J  Wasn’t high yaw-rate spinning in a circle also triggering?  Like if you did 3-5 “spins” in place it may lose (and then regain) heading?

 

From: drones-...@googlegroups.com [mailto:drones-...@googlegroups.com] On Behalf Of Robert Lefebvre
Sent: Monday, January 12, 2015 6:15 PM
To: drones-discuss
Subject: Re: [drones-discuss] Re: Acro Trainer Mode and Frame Reccomendation

 

No, it won't autoland... unless something changed that I'm not aware of.  Last time I flew it was maybe September.

[Robert Lefebvre]

Yes, absolutely, but yaw spinning is just a subset of rotations.  The problem is continuous rotation in any axis.  At least, this is as I understand was explained by Paul.    And I have had it fail from that.  I've also had it fail, doing this manoever... hard to even describe.  Like riding a half-pipe.  Hammer-head, into a hammer-heat, into a hammer-head.  Except instead of just a 180 rotation at the top, I was doing 2-3 piros.  But I think this was just the same cumulative effect of too many rotations too close together.

I really do wish I could shoot a good video of what the system is capable of.  It's hard to imagine a beginner breaking it.

I have wanted to go back and make a "pure acro" mode, which is basically what I had before.  No EKF, no DCM, just look at the gyros, integrated the errors, and that's it.  Then it can't fail no matter what you do.  The only problem with that, is if the user wants to bail-out to Stab, it could well bail-out upside-down, because the AHRS could be all screwed up and you wouldn't know.

[Josh Welsh]

See you’re doing piros on top and I’m just breaking in to hovering upside down and then hammering down again.  Piros.. ha.  That’s for people with better thumbs than me!

 

Come to San Diego for a visit, I’ll help you get a video J

[Robert Lefebvre]

I don't know if I'd call them piros... I'm not really that good.  It's actually a fairly simple manoever.  It's not much harder than a hammerhead in an airplane.  You just hold yaw longer.  But the acro controller does such a good job holding the roll and pitch targets, it's easy.  Just hold the yaw, release at the right time, pull back, collective, climb vertical, neutral collective, yaw again.  Pretty easy really.  You could probably also do it in a quad, sort of.

The biggest problem I have with acro on quads, is the zero throttle where it stops stabilizing.  

[Brian DeBusk]

I'm guessing continuous rotations on any axis would build-up the I-term of the PID calculation and leave you with a controller determined to over-compensate trying to stop the rotation (and a very confused EKF).  Your only hope would be to stop the rotation and let the accelerometers dominate the EKF estimate and re-stabilize.

I need to dig-into the APM EKF more, but that's how a home-grown one would react.

Brian

[Robert Lefebvre]

The PIDS really don't cause any problem at all.  With 3.1, when Acro was a simple algorithm reading on the gyros, this problem didn't exist.  The problem right now, is that the Acro control is utilizing the AHRS, and the AHRS is not suitable for this operation.

[Leonard Hall]

I have not seen anything to suggest that there is any problem right now. I have also not seen anything to suggest that AHRS is not suitable for this application.

As part of my standard testing I perform 5 minutes of consecutive single and double roll's and flips, with no delay longer than needed to maintain altitude. I don't see any problem with the position feedback from the AHRS when I do this test.

The AHRS may start to mess up with extremely fast and extended rotations but this is very hard to achieve in copters. Even in this situation. The stabilized correction using the AHRS has a limited correction that is normally set much smaller than the maximum rotation rate the pilot can command. So the pilot won't see this until they let go of the sticks and the copter continues to move slowly. In this case the pilot is still able to control the copter. The danger in this situation is if you switch back to stabilize. In any case, acro is just as capable and stable as stabilize.

[Robert Lefebvre]

Leonard and I chatted about this, and I think the issue is really more about the AHRS (DCM or EKF) than it is about the Acro mode performance, or comparison of current AHRS-Acro vs. previous gyro-only-acro.  My statements about sub-optimal performance, are really more about what happens when you switch from Acro to Stabilize, and are expecting the copter to come out perfectly level.

I am aware of several helicopter flybarless controllers, which have "bailout" features (aka: Self Level).  I generally only hear glowing reviews about these, the only time they are said to fail is if vibrations are too high.  And I know that helicopters are often used in extreme acrobatics.  I assumed that, putting 2 and 2 together, that you can do whatever the heck you want with these, and bail out to level without issue.

But upon further research, I'm not sure that's the case.  I really can't find any examples of radical acrobatics, followed by a switch to bailout.  The most aggressive example I can find is this one:

www.youtube.com/watch?v=hqYxnLe6ajo

That's not really even that aggressive.

I have had cases where bailing out to Stab resulted in 15-30 degree errors from level.  But this was only with DCM.  If the EKF is working, it's always virtually perfect.  And when DCM gets messed up, it's after doing some flips at much higher rates than shown in that video.

Now, starting about 2:15, there are some complicated manoevers, and the system works.  They are complex manoevers with rotation in all axes simultaneously, but they are not high rate.  I don't know if our system would be able to handle that or not, as I can't fly like that.  I also don't know if that system would have a problem bailing out after several very fast rotations.

In fact, I suspect that our system might even handle some manoevers, such as a Hurricane, better than they would, because we have GPS feedback that is able to tell the system that it's flying in a circle.  I can't imagine these other systems would bail-out properly after several hurricane rotations, it seems almost impossible. 

So, not to judge our system unfairly harshly, I think more investigation is needed to see how we truly stack up.  Maybe I'm expecting too much.

[Brian DeBusk]

Josh:

What battery are you using in your Cinetank?  I'm trying to find something in the 4000 mah 4S variety and they're all just a few mm too wide or tall (if you believe the battery compartment specs).

Brian

[Josh Welsh]

@Brian

Usually I’m using the A-Spec 3700.  I don’t need to, but, I bought three of them and they work great.  Prior to that it was Nanotech 3300 or 4000.