Dual-motor tailsitters

kd0aij · April 13, 2017, 6:40pm

@tridge Here’s an interesting event: uncommanded pitch up of 40 degrees, to which I responded with a small throttle increase and a lot of down elevator. But it appears that motor current increases in proportion to pitch angle. Leonard’s patch isn’t in master yet, is it? In any case, the error was corrected within about 1 second. I’m guessing the deviation was caused by wind or turbulence, possibly causing a yaw demand which saturated the elevons.

log: https://drive.google.com/open?id=0Bw3digSMQXDuN254dUEzVUUxMFU

adding yaw P term to plot, it seems that the yaw demand precedes the pitch deviation:

tridge · April 13, 2017, 9:10pm

quadplane inherits the hover throttle code from copter, but it doesn’t yet have automatic learning of the hover throttle. So the hover throttle setting in Q_M_THST_HOVER does work, but we need this function to make it learn:

github.com

ArduPilot/ardupilot/blob/master/ArduCopter/Attitude.cpp#L106


      
              }
          
              return desired_rate;
          }
          
          // get_non_takeoff_throttle - a throttle somewhere between min and mid throttle which should not lead to a takeoff
          float Copter::get_non_takeoff_throttle()
          {
              return MAX(0,motors->get_throttle_hover()/2.0f);
          }
          
          // set_accel_throttle_I_from_pilot_throttle - smoothes transition from pilot controlled throttle to autopilot throttle
          void Copter::set_accel_throttle_I_from_pilot_throttle()
          {
              // get last throttle input sent to attitude controller
              float pilot_throttle = constrain_float(attitude_control->get_throttle_in(), 0.0f, 1.0f);
              // shift difference between pilot's throttle and hover throttle into accelerometer I
              pos_control->get_accel_z_pid().set_integrator((pilot_throttle-motors->get_throttle_hover()) * 1000.0f);
          }
          
          // rotate vector from vehicle's perspective to North-East frame

it should be pretty easy to add to quadplane.
We’d need to make sure it only learns while in a steady hover. If we tried to learn in a transition or assisted flight we’d end up with some really wacky values.

tridge · April 13, 2017, 9:14pm

that depends on your Z controller settings. Just raise Q_ACCEL_Z and Q_VELZ_MAX and it will jump pretty hard in QHOVER. Also set hover throttle correctly. Some vehicles will need 0.7 or 0.8.

tridge · April 13, 2017, 9:17pm

it is used in quadplane, as it uses the same position and attitude controllers. Look for get_throttle_hover() in the code. It is used in many places, but the most important is this one:

github.com

ArduPilot/ardupilot/blob/master/libraries/AC_AttitudeControl/AC_PosControl.cpp#L497


      
                  _accel_target.xy() *= decay;
              }
          
              init_xy_controller();
          }
          
          /// reduce response for landing
          void AC_PosControl::soften_for_landing_xy()
          {
              // decay position error to zero
              if (is_positive(_dt)) {
                  _pos_target.xy() += (_inav.get_position_xy_cm().topostype() - _pos_target.xy()) * (_dt / (_dt + POSCONTROL_RELAX_TC));
              }
          
              // Prevent I term build up in xy velocity controller.
              // Note that this flag is reset on each loop in update_xy_controller()
              set_externally_limited_xy();
          }
          
          /// init_xy_controller - initialise the position controller to the current position, velocity, acceleration and attitude.
          ///     This function is the default initialisation for any position control that provides position, velocity and acceleration.

tilt · April 13, 2017, 11:19pm

Hi, tridge, got google hangout up and tried you, using my g mail, wifes not up on what you mean, gave me that look. so you want to try contacting me ?

tridge · April 13, 2017, 11:33pm

sure, what’s your google email address and I’ll contact you now
send it as a PM if you prefer
mine is tridge60@gmail.com

tilt · April 14, 2017, 12:08am

ryonika62@gmail.com, sorry new to this

kd0aij · April 14, 2017, 12:30am

@tridge Tried qstabilize mode at the field, but the wind was too high for the Stryker. Wasn’t able to command pitch greater than about 30 degrees, and the wind was carrying it away.
Which parameters determine the min/max pitch range? increasing LIM_PITCH_MIN/MAX seemed to have no effect.

Log: https://drive.google.com/open?id=0Bw3digSMQXDuZUtjX2hpZUk5eWM

Takeoff is at 16:20:29, flight lasts about 20 seconds with one uncommanded 360 degree (body-frame) roll.

I was quite pleased with the wind handling capability of the Convergence; takeoff and landing in hover mode required very little extra effort in the wind (avg. 12mph, gusts around 20mph). There is essentially no pitch change when heading directly upwind, motor tilt provides the necessary airspeed and wind penetration capability is quite high. Landing on any heading other than upwind would be difficult, though. So I think I’ll put a PixRacer in the Convergence before I build another tailsitter.

tridge · April 14, 2017, 3:52am

the limit in Q modes is the minimum of Q_ANGLE_MAX and LIM_PITCH_MIN/MAX
I see you pushed LIM_PITCH_MAX to 80 degrees. Might want to reduce that a bit [quote=“kd0aij, post:254, topic:15302”]
So I think I’ll put a PixRacer in the Convergence before I build another tailsitter.
[/quote]
great!

kd0aij · April 14, 2017, 3:52am

I wasn’t able to find any code that read the value of Q_M_THST_HOVER, only MOT_THST_HOVER showed up in my search.

tridge · April 14, 2017, 3:53am

for those following along, Rick and I had great fun on a G+ hangout and we sorted out his issue. The problem was that he had loaded the parameters from the PixRacer into the Pixhawk, including the BRD_TYPE parameter. That caused BRD_TYPE to be set to 4, which caused the startup on the Pixhawk to fail.

kd0aij · April 14, 2017, 4:11am

@tridge It looks like I’m going to need 10 degrees of up trim for FBWA mode. What would be the best way to accomplish that?

tridge · April 14, 2017, 4:32am

do you also need it in hover?
If you only need it in forward flight, then with current code you’d need to just rely on the integrator fixing it in either hover or forward flight. Right now the elevon trim is the same in both. If you have a nice high I term it will learn it quite quickly, but that may not be good enough.
We could add a parameter for it. Maybe Q_TAILSIT_PTRIM ? That could be a -1 to 1 parameter for pitch trim in hover.

tridge · April 14, 2017, 11:48am

While working with @tilt today we found the change I made to allow elevons and motor tilt to operate while disarmed in Q modes didn’t work. I pushed a fix today. That should make it much easier to do pre-flight testing of control in Q modes.

lorbass · April 14, 2017, 12:43pm

Yes. it didn’t work, so far. Thout it was my missunderstanding.

And does the Param Q_M_THST_HOVER work?
See post of:

tilt · April 14, 2017, 2:14pm

the fix was to load rover then back to tailsitter,he also fixed every thing else I could think of, amazing skill set, thank you Tridge

Rolf · April 14, 2017, 3:53pm

We plan to build a vectored thrust tailsitter. Some questions before the construction of the tilt mechanism: Which maximum angle is required for the tilt amechanism of the front motors ? Are 45 degrees in each direction enough ? Is an asymmetrical deflection possible ?

Rolf

Walter · April 14, 2017, 4:15pm

Hi Tridge,
“Mozart” now flies perfectly and I have only to do “routine work”.
That’s why I just for fun built a delta. Base of the Might Mini Arrow by E-Flite.
But no kit. The surface is identically constructed, the fuselage is somewhat more spacious.
The aircraft flies completely manually controlled very well. Now we want to convert the aircraft to a tailsitter. I have read many pages here, but I still have a few questions:
What configuration is better for this delta? Vectored or not?
Are the 120 degrees of a servo for the vectored tailsitter ? - 0-90 or + 30 / -30 degrees for oscillation-
How does the motor position relative to the center of gravity (in the longitudinal axis) affect the stability behavior?
Regards
Walter

@ Rolf,
sorry, but we can make also +30/-30 degrees for oszillate and 90 degrees for horizontal start.
Because the servo has 120 degrees max. (If Tridge confirms my question)

kd0aij · April 14, 2017, 4:17pm

@lorbass Looking at my log from yesterday, with Q_M_THST_HOVER set to 0.7, it seems that hover is still at about 75% throttle. @tridge says the parameter is implemented, so perhaps I don’t understand what it’s supposed to do, or maybe it needed a reboot. I changed that parameter at the field, and may not have rebooted afterward.

tilt · April 14, 2017, 8:14pm

Hi,if this is at all helpful