Ryerson flying tomorrow Sat Nov 5

Started by Frank v B, November 04, 2016, 07:28:04 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Frank v B

We are going to attempt another flight with the Ryerson Autonomous Project tomorrow, Saturday, Nov 5.  12-2:30 pm.
Fingers crossed. :D

Frank
"Never trade luck for skill"

sihinch

I'm hopefully meeting Ed there, similar timing ish.

Frank v B

What a beautiful day.... but why does the wind always have to swirl over the trees behind us (westerly breeze).  Let's have it directly down the runway for once.  The Ryerson team will hopefully post the videos (3 flights!) and some of the data logging. 

Frank
"Never trade luck for skill"

altonyeung

Great flying Frank! Your inputs are so important for us to improve our plane. I can say the plane has departed the bottom of the "brown pants scale" and slowly become flyable. On the last flight, we performed some tuning and it was a huge milestone for the project. Our plane should now be only a few flights away from autonomous flights. We can then perform orbiting circuits out in the field and gather useful data.

Here are the onboard camera footages of all three flights. Thank all TEMAC members for the kind support.


Flight 1 (Extended vertical fin. Differential aileron setup)
https://www.youtube.com/watch?v=CJE4SEE3o6U

Flight 2 (Added weight to correct CG. Increases differential aileron down throw)
https://www.youtube.com/watch?v=vAcgxHNVnro

Flight 3 (Control turning. Measuring pitch and roll response)
https://www.youtube.com/watch?v=uvScIsTjDeg

- Alton

Frank v B

Lemme Splain the videos:

flight video 1- purpose of the flight was to re-trim the plane after adding fin area and some differential throw on the ailerons (more up than down).  The plane was tough to handle for two reasons: i) there was not enough aileron authority to turn the plane.  With full aileron deflection the plane would only go straight until some rudder was used.  Then it would kick around.  ii) The ailerons were neutralized after the re-work whereas the plane flies straight with a lot of left trim (left aileron up).  Once the trim was added, it flew better.  Why the left trim needed?  We all know the purpose of the trim levers is to correct building errors.  The wings must be warped a little. ;D
The plane also flew like it was tail heavy.  (sensitive on pitch)

Flight video 2-  More aileron deflection was added and about half of the differential was taken out.  We also put about 3 oz of lead in the nose.  The plane flew a lot better and much less pitch sensitivity.

Flight video 3  Alton wanted to start programming the autonomous electronics.  We did about 3 circuits with full aileron deflection left and right for about 3 circuits.  The next three circuits was full up and down elevator to get the stabilization unit learn the pitch sensitivity of the plane.
No Andy, the pilot was not drunk.  He was just flying the way he was asked to fly.

The next flight will be probably be next Saturday and it will be to test the autonomous system itself.  We accomplished it on the foam Bixler earlier in the season.  The objective is to scribe a flight plan on the computer screen that has a Google Earth shot of the field.  Whatever flight path is scribed on the screen will be what the plane will follow via GPS.  Any safe altitude can be picked.  If it starts off at 20 meters and then is asked to do the pattern at 30 meters, it will automatically ascend and loiter at that altitude.

My role:  the plane cannot take off and land*.  My job is to land it in as many pieces as it took off.  One and only one. Job security! 8)

Frank

* sometimes I can't either but don't tell Ryerson. ;)
"Never trade luck for skill"

Bill B

The data seems to back up what Frank said...
Looking at the first two flights which were full manual (Frank) flying, the nose was correctly pointed a few degrees into the wind at all times (seems the larger tail is helping).

I've attached a figure with data from the third flight, to help give a sense of what happens during an auto-tune. The first two charts are roll data. Chart 1 shows Frank's demanded roll rate (in degrees per second, shown in black dots) over the flight, and the actual roll rate, shown in blue dots. As you can see, at the beginning of the flight, the achieved roll rate has large error from the demanded. The auto-tune adjusts the value of "p" during the flight to try and have the achieved roll rate match the demanded. You can see that, as the 'p' value rises over the flight, the achieved rates become closer and closer to the demanded. However, the 'p' value is still rising by the end of the flight, meaning that more auto-tuning with roll is necessary (We should auto tune until the P value levels off).

Charts 3 & 4 show pitch data. It seems the achieved pitch rate never quite reached the demanded... this suggests that we require much more auto tuning on pitch.

Looking more into the auto-tuning procedure, the online instructions say to perform at least 20 full stick movements for both roll and pitch. Next flight we should do some more auto tuning, and then maybe look at manually fine tuning the flight parameters (things like maximum roll angle, maximum roll rates, etc.). Then the plane should be ready to do full autonomous flight (except take-off and landing... we still need Frank...)

Thanks again TEMAC,
Bill

altonyeung

Thanks Bill for posting the plot for us. Yes, we may need further tuning for roll and pitch. As we can see from the chart, the actual roll rates are almost reaching the demand rates which suggests the P-gain value is very close to a correct one.
For the pitch response however, I think the default 75 deg/s roll rate is just too much for the short-coupled aircraft with "not enough" elevator. We will look deeper into the actual elevator servo response and deflection angle. We should be able to come up with proper roll and pitch rate for this specific airframe before the next tuning flight.

We will post further updates regarding this flight in this thread.

Frank v B

"Never trade luck for skill"