I am trying to tune an SAB lizard turbine plane (10kg plane with 120N thrust) for auto runway landing. I also wish to do VTOL tailsitter hovering later on.
After doing autotune for roll and pitch, I tried an auto mission with circuit waypoints. The plane overshoots the waypoints and overspeeds through the turns. Here is the log.
After looking at the arduplane tuning recommendations, I plan to make the following changes
WP_RADIUS 40 → 100 (to start turning earlier)
THR_MAX 100 → 80 (to prevent excess speed. however, I need 90+% throttle for tailsitter Qhover, so I might need to raise it again if I do Qhover)
You should adjust scaling airspeed and retune as it is way off. (IIRC if you set it to 30m/s you will need to decrease rate PID values by 4 as control deflections are scaled by 1/V^2)
You should fix TECS tuning instead of limiting throttle.
Tuning looks like it could be made a bit more aggressive, especially in pitch as you are using very little of control surface range. (May be accomplished by tuning with correct tuning airspeed).
There is something wrong with your airspeed data and TECS is falling back to throttle based control. 13m/s airspeed at rest is pretty bad.
I am afraid tuning hover with that will be quite hard as you have little thrust margin and slowly responding. You will also need some sort of thrust vectoring and yaw control.
use ONESHOT_MASK to setup all the fast control surfaces as oneshot. That will gain you a bit of phase margin in the control loops
I suspect you will do better with roll and pitch PIDs dominated by P and D instead of FF. The autotune process produces a FF dominated tune
The main thing I’d recommend though is to find the closest aircraft to yours in the RealFlight user uploads and then get RealFlight working with your plane so you can do testing in a simulator. If you pay careful attention to the throws, weight distribution, thrust etc in the simulator you should be able to do a fair few tuning experiments there. You won’t get things exactly right, but I suspect it will help.
I haven’t managed to test out the changes for horizontal flight yet. But I have separate questions for vertical hover throttle control from past testing. I have thrust vectoring and 2 small edfs at wingtips for roll control.
I am launching as a tailsitter. At the start, the throttle takes a long time to increase (Qhover mode) from idle to 90% needed to takeoff. Is there a way to make this faster?
I reduced the vertical controller gains (Q_P_VELZ_P and Q_P_POSZ_P) to a very small value to cope with the delay in turbine thrust control, but it is still oscillating during hover. Any parameters I should tune to improve vertical control?
Any suggestions on how to test the hover to forward flight transition safely? What parameters should I focus on for transitions?
I managed to do another test flight and the plane tracks the waypoints much better now. Thanks for all your suggestions The plane altitude varies about 20m in the circuit though, not sure if that will be an issue for auto landings.
I am not sure which parameter changed made the difference though, since I changed a few at once
Main changes were:
Change scaling speed 15 > 40 and corresponding changes to PIDFF values (I tuned the values based on the old values and SITL and realflight simulations. I tried to redo autotune in flight but it did not complete and the controls felt acceptable)
Add one shot to control servos
Increase WP radius 40 > 150 and increase distance between waypoints
Small tweaks like (trim throttle 36 > 30, TECS_CLMB_MAX 5>15, TECS_SINK_MAX 5 >10, TECS SINK MIN 2 > 10, AIRSPEED_MAX 70>65, AIRSPEED_MIN 27>25, PTCH_LIM_MAX 30>20, PTCH_LIM_MIN -30>-25)
Here is the log. Anyway I would appreciate any suggestions for improvements Thank you!
The plane altitude varies about 20m in the circuit though, not sure if that will be an issue for auto landings.