Here is the log for the 1st Roll bench test. Throttle constant
P=0.4, I=1.0, D=0.01, Q_A_Ang_RLL_P=2
https://drive.google.com/file/d/1_n-6cDR2GAEnpc1D1PVB_fVMayZd4OLP/view?usp=sharing
that looks ok for a test hover.
Ok. Vehicle glueing now. Will try test hover tomorrow and report results.
Thank, William
try the Roll I at 1.5 if you find the roll unstable.
Hi, I tested the vehicle in hover today several times. It behaved well and their were no crashes. The vehicle was held in ground effect lifting at most 1 foot. Because of its close proximity to the ground, it touched the ground sometimes. The acceleration in the logs will give you an idea when it touched. The vehicle hovered the highest in the last 2 flights. Here are the logs for today:
https://drive.google.com/file/d/1AMo4AcV-on91pp4wnZB4zQApb4EamyrR/view?usp=sharing
https://drive.google.com/file/d/1stDvT8Q1UzIm-oO38INANn3CJm0V4xZM/view?usp=sharing
https://drive.google.com/file/d/1b5SFRaxgx397EfKvPNwWbk2WeWQ9QhZJ/view?usp=sharing
https://drive.google.com/file/d/1YvHSTdKG1MtgDMRS9Niq6TeM_Oc671pO/view?usp=sharing
https://drive.google.com/file/d/1oH86VDrZwvO7Xk_LPc8G6zTG64DYPSLu/view?usp=sharing
https://drive.google.com/file/d/1wDOBqWBb_a2oIlBYurCCmn29CsV0TDTm/view?usp=sharing
https://drive.google.com/file/d/1JSJhSMuOGnBbbMUNTT_aA7E6KfV5DSnr/view?usp=sharing
I should mention I accidentally had about 30% Right Elevon trim during all flight (residual from bench tests and forgot to reset). All other trims neutral and level was set in MP before tests. This is probably why the vehicle kept turning right a little.
Videos of first 6 flights:
https://drive.google.com/file/d/1_V2MW29ErSz7N0MKA1qklTM-qcdghiZF/view?usp=drivesdk
Video of last flight:
https://drive.google.com/file/d/1_Z04ZxUMaSzOqKGioAnw6EqWJaam19DY/view?usp=drivesdk
much more I gain, increase the roll I to 1.5 and pitch I to 2.0
Ok. But closer inspection of vehicle shows damage to back wheels that needs to be repaired before another test can be done.
change yaw I to 1.0 to stop it yawing. also check all your trims are centre or your going to get unwated movements.
I ran 4 experiments today for: Roll I=1.5 (1.0 before), Pitch I=2.0 (1.5 before), and Yaw I=1.0 (0.018 before).
Here are the logs:
https://drive.google.com/file/d/1b7pDsvuH8m5HIu_akdtnQg7fKMiuRCSG/view?usp=sharing
https://drive.google.com/file/d/1Fo4f2C2DJWom4ThXo-X1j579Zz1U2ZDS/view?usp=sharing
https://drive.google.com/file/d/1iuwDaLWTQRcBB-2zbWKND3lImpgS8e84/view?usp=sharing
https://drive.google.com/file/d/1MHHkpIN9UlSqZjn7w_vT4RHV7xRMxnCe/view?usp=sharing
Here are the videos for the tests. They are in the correct order, but one video may be missing:
Test 1:
https://drive.google.com/file/d/1birXIEqEmZ9F7h_50lHUpr1OfIJ_RRsg/view?usp=drivesdk
Test 2:
https://drive.google.com/file/d/1bqrBtoRqMnhoa-MrJuWK0erC2N0J6CXq/view?usp=drivesdk
Test 3 and 4:
https://drive.google.com/file/d/1btRb8CmvnUSFanOSKq7d1pZ6_dW_CMOz/view?usp=drivesdk
32.bin (2nd to last flight) Pitch tracking Despitch well
30.bin (1st flight) Roll and Desroll. Oscillation due to hitting floor
33.bin (last flight) – See 1.5 deg pp Oscillation in Roll (I gain too high?, but last time not well tracking , so make in between)
33.bin (last flight) – Too much gain and Yaw oscillation - need to turn down YAW I gain
29.bin (last flight of previous settings) – Yaw is tracking good
Notes for this test:
Roll I=1.5 (1.0 before), Pitch I=2.0 (1.5 before), and Yaw I=1.0 (0.018 before).
Conclude: Pitch ok, For next test will change: Roll I=1.3 (1.0 bad track, 1.5 Osc), and change: Yaw I=0.018 (0.018 track well, 1.0 Osc , )
Hi. I have tuned the vehicle the best I can using IPD tuning and ground effect testing. Here is the resulting final PID Values:
Pitch: P=0.3, I=2.0, D=0.005
Roll: P=0.4, I=1.2, D=0.01
Yaw: P=0.18, I=0.5, D=0.0
The Pitch, Roll, and Yaw seemed to track the desired stick commands. So I tested the vehicle yesterday outside and tried flying it higher out of ground effect. The overall behavior was about the same as the flight with PID tuning. Here is video and log of the test:
Video:
Log:
Here is the Video and Log of the previous outdoor test using PID tuning:
Video:
Log:
These were the PID values used:
Pitch: P=1.0, I=0.02, D=0.01
Roll: P=1.6, I=0.2, D=0.01
Yaw: P=0.18, I=0.018, D=0.0
My conclusion is that the IPD tuning did not really help and large ~50% Throttle changes still create large ~5 deg Pitch changes. I am also seeing ~1 deg Pitch Setpoint drift. So the vehicle is still very difficult to fly. Here is a picture of the Pitch (green), Despitch (red), and Throttle (blue) for the most recent test/crash:
I have done some testing in SITL using a similar in function Convergence RC model. The same changes in Throttle inducing changes Pitch is observed. But it is not observed in the stock mini Convergence that comes with RealFlight, which uses a different flight controller algorithm. Also I have developed my own RealFlight simulation of my full scale vehicle, which does not use a flight controller and flies very well, and shows no affect of throttle on pitch. These tests indicate that the flight controller in Ardupilot is not working well and is causing this problem. Please refer to the following link for more detailed information about these tests and discussion:
William
I have just tested the 1/6 scale model and the Throttle Pitch effect is also observed there, I see (~5 deg) pitch changes with large changes in throttle (60%), but it is hardy noticeable when looking at the vehicle.
But the effect seems to be more noticeable with the full scale vehicle even though the pitch changes are about the same. Perhaps I am just standing too close to the vehicle when I fly it and this magnifies the effects. The 1/6 scale uses the same motors and ESCs in front and back. All the props are the same diameter. But the props in front are 2 blade, whereas the back prop is 3 blade to give it more thrust. The back motor is running hot and I have had to replace it several times. So the model could use a more powerful motor.
Here is a calculation showing that offsets in the Thrust Throttle curves create a nonlinearity that makes the Pitch depend on Throttle:
To get rid of the Pitch dependency on Throttle we need to linearize the Thrust Throttle curves of both front and back motors, and also get rid of offsets in curve so the curves go through the origin.
Here is the Mathematica code for the Taylor Series Expansion about t=a:
Note if b1d1 ~ b2d2 then G2 becomes almost constant (G2= G1d1/d2) and not very dependant on t. So this condition will make the Pitch not very dependant on Throttle!
I am comparing the 3 Full Scale vehicle test flights:
- A good indoor flight with PID tuning (ie P larger than I term).
Pitch: P=1.0, I=0.02, D=0.01
Roll: P=1.6, I=0.2, D=0.01
Yaw: P=0.18, I=0.018, D=0.0
Video:
https://drive.google.com/file/d/1csnLLtug1sz9msU4BM_flKp9pHXlquIa/view?usp=sharing
Log:
https://drive.google.com/file/d/1s8Q2lW3JI-FuHghaXTEG955r7Wb6tA5l/view?usp=sharing
- An outdoor crash with same PID settings
Video:
https://drive.google.com/file/d/1XF2af9BRVl0FTPep8M5Kbumt_HAGXxAh/view?usp=sharing
Log:
https://drive.google.com/file/d/1sIP8zdrSVs5vN-fAD5S_UVAoOIKZN91y/view?usp=sharing
- An outdoor crash with the new IPD settings
Pitch: P=0.3, I=2.0, D=0.005
Roll: P=0.4, I=1.2, D=0.01
Yaw: P=0.18, I=0.5, D=0.0
Video:
https://drive.google.com/file/d/1gxa5mDEFPype01eP6R1W1ZBdAjq6iy_z/view?usp=sharing
Log:
https://drive.google.com/file/d/1kSqEx0E3emf8R_CD-Wx1iPFjBJELHhlx/view?usp=sharing
- Signals from Flight 1 :
Pitch (red),
DesPitch (green
Throttle (blue)
- Signals from Flight 2:
Pitch (red),
DesPitch (green)
Throttle (blue)
- Signals from Flight 3:
Pitch (red),
DesPitch (green)
Throttle (blue)
Comparing, all 3 flight, one can see that the original PID tuning was better. The DesPitch to Pitch was faster and the Throttle to Pitch effect was about the same for the last 2 flights (~5 Deg change in pitch for ~50% change in Throttle.
Comparing the 1st 2 flights with same PID settings. One can see that if the changes in throttle are more gentle (ie have less slope), the Pitch does not change much
That’s amazing! Congrats to the team working on this kind of contribution with academic background and hands-on experience.