Local Motors/Airbus VTOL UAV takes to the skies!

3689705547?profile=original

Hey everyone, 

I wanted to announce the successful first forward flight with the Local Motors/Airbus VTOL UAV. Powered by a Pixhawk 1 with Arduplane 3.7.1, weighing in at 54.85lbs!

I was called in three weeks ago to help them design, build, and fly all before 11/30/2016.  I showed up at the shop and they had nothing but some ESCs, Motors, and concept art. Well on Wednesday we made the attempt to transition, and it came back in one piece.

The full report and video of the flight can be found here:

https://launchforth.io/localmotors/airbus-cargo-drone-build/discussion/topic/flying-drone-final-build-test-flight/62079/

 

3689705620?profile=original

E-mail me when people leave their comments –

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

Join diydrones

Comments

  • So Joe - does this mean you've now flown the smallest AND the biggest ardu-QuadPlane?
    Well done btw
  • Hey Tridge,

    Excellent analysis, spot on with every assessment:

    We would get a BAD AHRS message repeatedly, even after re calibrating the accel but only on the last day when we got to the field. We only had one day to fly and all other indicators seemed normal so we went ahead with it.

    You are right about the airspeed sensor, I opted out because I didn't want things to get complicated. I would prefer flying with one but basically this was just a cruise around the pattern and back to get some data points.

    Vibration looked good in the hovers as well, but there was some flexing in the frame, specifically the tail. It wanted to rotate a bit as we powered the rear motor. Unfortunately, we couldn't do anything to really reinforce that area because we were down to grams trying to get the CG forward. 

    We knew there was a yaw problem, in fact it was way worse until we added a few washers under the motor mounts. We simply ran out of time to get most of this right but there idea was to fly no matter what.

    We had tried to spec out a different pusher system, but we simply ran out of time between ordering something and mounting it to the airplane.  We intentionally set the quad motors to stay on a little longer than hopefully needed so that it was always under control. We had very little documentation or understanding of the aircraft so we used our best guesses as to the speed and gain sets for fixed wing.

    Everything you said completely makes sense! Again, that was an excellent analysis.  They pretty much told us this thing had to fly no matter what, so we did what we could with very limited time.  I basically told them that if they want to move this to a phase two, we can do wonders for this program with lessons learned.  Thank you very much for your work here, I will be sure to pass that on to the right people!

    PS I never got to say congrats on the OBC 2016! Awesome work!!!

  • Developer

    @Joseph, I had a bit of a look at the logs.

    A few setup things first. The logs show ARMING_CHECK=0, so it didn't do any pre-flight arming checks. I'd highly recommend setting ARMING_CHECK=1. It didn't cause a problem for this flight, but is worth fixing.

    I also notice that it had no airspeed sensor enabled. While we do support flying quadplanes with no airspeed sensor, it would be a good idea to add one. I'm guessing you were trying to just keep things simple for the initial flights?

    Vibration was generally OK with hover under 2G, but there were some periods during the transition where it got to about 6.5G, and even had a few clipping events (which means it was over 15.5G momentarily). Such a large range of vibration is unusual. Vibration was much worse during transition than hover. You didn't actually reach Q_ASSIST_SPEED for Q_TRANSITION_MS so the vertical lift motors never fully disengaged, which makes it a bit hard to tell if it is the forward motor that is causing the vibration or not.

    The attitude and climb rate control were not too bad during hover. That is actually a bit surprising as the gains on the quad attitude controllers were really low for such a large aircraft. It was a bit sloppy, and I think you could push the P and I gains up a fair bit. Don't push the D up much though. In vertical flight it seems to be slightly tail heavy, but not enough to worry it (the pitch integrator got to about 5% nose down).

    The yaw was a bit off in hover. It is tending to yaw right, and the yaw controller is having to waste a significant amount of power compensating.

    trans2-hover-motors.pngThe easiest way to fix this would be to angle the right motors slightly forward (we had this in our quadplane and we put 1mm washers under just the rear of the motor mounts on one side to fix it). Ideally you'd get it so hover in zero wind puts the same output to all motors (or as close as you can get).

    The two biggest problems are with pitch control during transition, and not enough power from the forward motor during transition.

    The forward motor was close to full throttle for 15 seconds but the aircraft only managed to make it to 14.5m/s, and that was with a bit of nose down. The Q_ASSIST_SPEED was 15, so it had to get past that speed for 5s before it will stop the vertical motors completely. Also are you sure 15 is high enough for a plane like this? I need 18m/s on my 2.7m quadplane (weight 15kg) to transition with a good safety margin. Your plane is a fair bit bigger so unless the wing loading is much lower I'd expect it to have a higher stall speed. You want your transition speed a significant margin above stall speed (say 3m/s or more above stall speed).

    trans2-thr-pitch.pngMy guess is the VTOL motors are causing too much drag while they are spinning for the forward motor to get the aircraft up to speed. That is a common problem in quadplanes, and can mean you need a more powerful forward motor than you might otherwise need. I notice in your reply to Rob that you said you needed to reduce the size of the fwd motor to get the CoG right. I think you'll need to find some way to solve that. Maybe even angling the VTOL motors fwd a few degrees will help? Then at least you won't be fighting them quite so much on transition.

    The second big problem is pitch control during transition.

    trans2-pitch.png

    that is pretty bad, with pitch going all over the place. Let's look at what the fixed wing pitch PID controller is doing:

    trans2-pitch-pid.pngFirst off, you can see its building up a fair bit of negative integrator. It saturates at the full 44% PTCH2SRV_IMAX, which is a lot of down elevator on the vtail. That means that either there is a significant up pitch moment from the vtol motors during transition or the CoG in fixed wing flight is too tail heavy. You won't really be able to tell which it is till you can complete a transition and get some clean flight time logged as a pure fixed wing.

    Because the pitch is a fair way off (and too nose high) the VTOL motors are pulling backwards for a lot of the transition. That makes the job of the forward motor really had as it tries to overcome the VTOL motors.

    There are some odd things going on with the VTOL motors during transition. The front left motor goes to zero during part of the transition:

    trans2-motors.pngthat is quite bad. You can see the motors are smooth (but with some yaw discrepancy) in QHOVER, then go wild during transition. We see periods where C7 goes to zero (thats the front left motor) and other periods where C5 goes to zero (thats the front right motor). It looks like engaging the fwd motor is making the yaw effect worse?

    You also have a power problem of some sort.

    trans2-power.pngthose Vcc values are much too low. At one point it reaches 4.2V, which is getting towards dangerous levels. How are you powering the Pixhawk? Is there something drawing too much current off one of the peripheral ports? The voltage on the servo rail also varies much more than it should, especially in hover when the servos won't be under much load. It looks like voltage/current sensing on the main batteries isn't enabled?

    I'm glad to see you got it flying and look forward to seeing some logs of it really stretching its legs in future posts!

    Cheers, Tridge

  • Well done Joe. Glad it Arduplane got it flying in the end. ;-)

    Have you tried forward motor assist and wind feathering yet? I think with a bird this size this will become crucial control methods in wind, due to the large wing area exposed to wind.

    Also have you added some forward tilt on those hover motors? (we use about 7degrees) This will help hover attitude control, in particular when using forward assist, to stabilise it in wind as the wings at a pitch up AoA will produce some stabilising lift, rather than the typical negative lift produced by pitching forward in quad hover. (Which you need to move forwards without forward assist.)

    If dialled in this should improve hover performance and increase your hover flight time in wind. To the point you need to be careful in high wind, as it will not want to descend if flown in wind just under wing stall speed. Note that typical tip stall are prevented by the attitude stabilisation of the hover motors and hignher AoA.

    You probably already figured this out, but note that with a QP the pusher motor and prop can be configured only for cruise, as it is typically not required for static thrust at launch because the QP can maintain lift on takeoff. This means if you really need to, you should be able to get away with reducing the forward motor and prop size and going to a higher kV which might help get the CoG right. The motor/prop combo should be the right pitch for the cruise speed and should be limited to the amount of thrust required at cruise, ignoring launch. (this typically ends up being a smallish high pitch prop). Being able optimise thrust to cruise also significantly improves overall efficiency.

    Keep us up to date how you go and we'd love to see some videos in flight.

    Regards

  • And a good set of thumbs!  The pusher motor was certainly under powered, we couldn't keep the weight under 55lbs and get the CG right without switching to a smaller pusher motor. One of the drawbacks of this design is the large moment arm of the pusher motor which makes CG difficult and also makes pitch corrections in quad mode struggle because of the inertia.

  • Congrats Joe. The aircraft looks much better for sure. Glad to see Ardupilot working as I knew it would, as I've also done some consulting on a very large quad plane like this. It takes sophisticated control algorithm to fly an aircraft like this.

    Does it need more power for full transition to forward flight?
  • Hey Tridge, couldn't have done it without your contributions! The logs are posted here:

    https://launchforth.io/localmotors/airbus-cargo-drone-build/discuss...

    I recruited some reinforcements to help with tuning on the last two days. Couldn't have done it without them either!

  • Developer

    great stuff! If you post the logs I can have a look and see if I can offer some suggestions for tuning

  • Basically they needed help, and I just repeated what I had already tried and had success with. Though it wasn't how I would have went about designing the aircraft, when I showed up a few weeks ago they were committed to it at that point. You can read more about how I contributed to the project here:

    https://launchforth.io/localmotors/airbus-cargo-drone-build/discuss...

  • If I remember correctly they were using PX4 as the flight stack. You now have ArduPilot, what caused the change?

This reply was deleted.