Thread Tools
Oct 12, 2008, 11:03 AM
Registered User
dmgoedde's Avatar
Thread OP
Atto has several layers of process control going on. The overall problem is broken down into a bunch of linear pieces, but together has a non-linear and organic feel. A lot of the navigation code and schemes have not needed an update since I first wrote them 1.25 years ago.. the crabbing for example. It is a natural result of how the navigation PID works.

Upside down I have thought of. It would be as simple as having a couple signs be flipped. It should work just fine, though I haven't tried it. I do know that certain ridiculous conditions don't seem to matter, such as grossly mis-balanced plane. The 16 ounce "Python" was quite tail heavy for several months and I never realized it until one day I took it to the hobby shop to see what larger LiPo could be put into it. The balance point was halfway back from the LE! The Ultra Mini Stick was balanced, then I rubber banded a 3.5 ounce camera to the left side of nose... autonomous flight was fine, RC required more trim than available so I had to hold the sticks "up" and "right" to go straight and level.
Sign up now
to remove ads between posts
Oct 15, 2008, 06:10 AM
Registered User
hoysome's Avatar
so.. the thread is 1/2 way off the first page, and no chitter chatter in 3 days. How is the release coming? What about AttoRTL?
Oct 15, 2008, 07:23 AM
Silicon/poker/chocolate chips
Chip Geek's Avatar

Video with large cross wind


Quote:
Originally Posted by dmgoedde
https://www.rcgroups.com/forums/show...01#post8223143
Video #11

This is roughly 30 seconds of autonomous flight on a 700m run at 150m AGL, airspeed 59 km/h, wind speed 36-40km/h! Road is East/West, wind from South. Plane crabs a stead approx 30 degrees to stay on course. Video is relatively stable given that the plane is being used at edge of its flight envelope.

50Hz attitude control of this 20 ounce elevon flying wing is using airpseed based gain scheduling, and optimized Kp and Kd.
That is very impressive! I would have liked to see what happened after the way point turn. Or more specifically, what would happen if that way point caused a 180 degree turn, then how does Atto handle the fact that the apparent wind (relative to the airframe) has now shifted 180 degrees. Does Atto realize that the wind is now coming from the left and crab that direction or does it still try to crab to the right for a while and get blown off course for a few seconds.

I guess a picture of the path for that flight would answer my question.

I would also love to see a flying pizza box. Just be careful about structrual integrity. The last time I checked, pizza grease and cardboard are not considered a strong composite structure.
Oct 15, 2008, 11:32 AM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by Chip Geek
That is very impressive! I would have liked to see what happened after the way point turn. Or more specifically, what would happen if that way point caused a 180 degree turn, then how does Atto handle the fact that the apparent wind (relative to the airframe) has now shifted 180 degrees. Does Atto realize that the wind is now coming from the left and crab that direction or does it still try to crab to the right for a while and get blown off course for a few seconds.

I guess a picture of the path for that flight would answer my question.

I would also love to see a flying pizza box. Just be careful about structrual integrity. The last time I checked, pizza grease and cardboard are not considered a strong composite structure.
Chip geek: Atto will retune the crabbing to new conditions in approx 5 seconds after the 180 degree turn. I didn't show a plot or continued video of the leg into the wind... because it is not pretty! It works and navigation continues along the path, but what you get is the same as experienced even with very expensive autopilots (or what I surmise from reading their manuals)... a zig-zag path.

Atto was being flown in a wind that was 66% of the cruise speed. The MicroPilot MP2028 manual states:

"Check that windspeed does not exceed 50% of cruise speed" as the last item in the pre-flight checklist.

So, based on my and Beta tester results, I can vouche that Atto will keep your plane safe at even >60% windspeed conditions, but that is far from ideal. The point of the video was to show Atto tracked to a WP in 'impossible' wind conditions, and kept the plane pretty stable. Atto did finish the course and entered holding pattern after the flight. It was a total of 4km at 150m AGL, and those 15 minutes of flight consumed every morsel of battery capacity due to battling upwind legs.
Oct 15, 2008, 12:31 PM
Inherent Tinkerer
Dean,

Does using impact air pressure to measure airspeed with a pitot tube have any impact on the max wind speeds that the plane can be flown in?

Jimmy
Oct 15, 2008, 05:07 PM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by jtprouty
Dean,

Does using impact air pressure to measure airspeed with a pitot tube have any impact on the max wind speeds that the plane can be flown in?

Jimmy
Jimmy - I have a feeling it might, or should allow the plane to fly in much higher winds... here is why:

Atto generates a firehose of data... 47 parameter columns at 5Hz in the current V1.7 testing (you wouldn't beleive all the stuff that can be modeled with this). I am going back over mountains of LOG data, and calculating and modeling special navigation that can make corrections for how navigation works depending on windspeed and vector of plane ground path with respect to wind vector. Real-time estimation of wind speed magnitude and vector is possible, and I suspect from there it is possible to correct how the navigation routine calculates roll angle targets depending on wind. If it works, result would be perfect ground path tracking under more extreme wind conditions. the system already works great under strong winds up to 50% of flight speed, but from 50%+ it gets difficult, or when plane ground speed drops below 10-15km/h.
Oct 15, 2008, 05:32 PM
Registered User
Quote:
Originally Posted by dmgoedde
Jimmy - I have a feeling it might, or should allow the plane to fly in much higher winds... here is why:

Atto generates a firehose of data... 47 parameter columns at 5Hz in the current V1.7 testing (you wouldn't beleive all the stuff that can be modeled with this). I am going back over mountains of LOG data, and calculating and modeling special navigation that can make corrections for how navigation works depending on windspeed and vector of plane ground path with respect to wind vector. Real-time estimation of wind speed magnitude and vector is possible, and I suspect from there it is possible to correct how the navigation routine calculates roll angle targets depending on wind. If it works, result would be perfect ground path tracking under more extreme wind conditions. the system already works great under strong winds up to 50% of flight speed, but from 50%+ it gets difficult, or when plane ground speed drops below 10-15km/h.

Dean,

Why does it get difficult when the plane ground speed drops below 10-15km/h?

Thanks,

Chris
Oct 15, 2008, 07:15 PM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by airmcn_3
Dean,

Why does it get difficult when the plane ground speed drops below 10-15km/h?

Thanks,

Chris
It is not heading error... you can see in a simple 7km/h walking test that heading is accurate and sensitive.

GPS is an Earth-based reference, but the plane is moving in a mass of air that is not stationary relative to the Earth... two frames of reference going on, and the plane is stuck between them. Living in the air and pushing off of it, but GPS reference is Earth-based. X amount of roll angle should yield Y amount of turn rate. But now if plane is moving into the wind, then rolls to say +30 degrees, its ground-based heading changes very rapidly as it turns to go with the wind. In still air ground and air frames of reference are very similar.

What I was discussing with Jimmy above is possibility of having Autopilot estimate wind speed and heading, then using this extra info to optimize navigation for better performance in high wind. Nonetheless, flying when wind is > 50% of plane airspeed (as a general line in the sand) is creating problems.

I tell Beta testers, that you must always consider flying conditions relative to the plane. It is a bad idea to fly a plane in conditions it is not designed for, no matter 50Hz attitude control or not. You can't overcome a bad physical situation entirely with advanced process control. A high wing polyhedral Miss2 is completely unsuited for gusty wind at 20 km/h, because airspeed is only 27km/h (based on my flights). Another reason beyond flight speed is the high wing trainer center of drag is up high due to the wings, but center of gravity is down in the fuselage... having them not aligned will make the plane get tossed around in gusty wind. That is a big reason I like the flying wing, and the old "Python" plane. Their wings are straight no dihedral and so center of drag and gravity are lined up well. Gusts of wind only buffet the plane translationally, not able to apply much torque to affect pitch or roll.
Oct 15, 2008, 08:21 PM
Registered User
dmgoedde's Avatar
Thread OP

Old new video uploaded..


https://www.rcgroups.com/forums/show...01#post8223143

I realized I had this old 1.7MB video from 11 months ago... this is old old AttoPilot on the Miss2 with 7 MegaPixel camera Olympus FE-230 in highest res video 640x480 pixel. Video shows what a cheap 32 ounce UAV is capable of... perfect clear and stable video, triggered at waypoints. Video like this can easily be routed through a wireless transmitter.

In this flight, 9 or 10 waypoints, camera video of 10 seconds at each waypoint was triggered by AttoPilot. This particular waypoint was directly over my friend's old house, as intended. Google Earth is only about 5 m "off" where I live, and the 'modern' GPS certainly does NOT have large 10m CEP like I heard claimed in another thread. Maybe it is because I am using truly modern GPS, not 2004 cutting edge SirfStarIII.

Dean
Oct 16, 2008, 01:16 AM
Registered User
Onboard video is SO much better than transmitted video it's crazy.
Oct 16, 2008, 10:57 AM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by carguy84
Onboard video is SO much better than transmitted video it's crazy.
Yep. Problem with the 20 ounce flying wing is 1) 3.5 ounce camera will really impact flight performance, 2) no good place to mount the camera in terms of balance or aerodynamics.

The video camera is very tiny, and is all I can manage to cram onto the flying wing. I am making a new small plane that can take a real digital camera, same 36" span elevon flying wing but short fuselage slung under with pusher motor in the back. Will be able to slide the wing to get right balance no matter the payload. Will use a lot stronger motor, aiming for max arispeed of 110 km/h like the Python. The python 26" plane could fly in any wind I ever encountered.

And - the music on that last video was not my first choice... it just happened to be on there from when I e-mailed the video the the guy that used to own the house.
Oct 17, 2008, 12:21 AM
In case you didn't notice... this thread has reached 100 pages. Nice job Dean!



Oct 17, 2008, 10:37 AM
USA
d_wheel's Avatar
I was looking back at messages from about a year back and noticed something I hadn't caught before. It stated that you were using the throttle for altitude control, and rudder for waypoint sequencing. The fact that you are using the latest version on a flying wing type airplane seems to indicate that is not the case now. Just wanted to make sure.

Later;

D.W.
Oct 17, 2008, 12:16 PM
Registered User
dmgoedde's Avatar
Thread OP
Quote:
Originally Posted by d_wheel
I was looking back at messages from about a year back and noticed something I hadn't caught before. It stated that you were using the throttle for altitude control, and rudder for waypoint sequencing. The fact that you are using the latest version on a flying wing type airplane seems to indicate that is not the case now. Just wanted to make sure.

Later;

D.W.
Ha ha!!! You got that right... oh my that was sooooo 2007.

Back then system needed plane with positive pitch response to throttle for altitude control, and altitude was only by GPS. Rudder deflection was based on heading error. It is really the very simplest autopilot possible. There is absolutely NO comparison between then and now.

Right now:
1) throttle control airspeed via pitot tube sensor
2) altitude is control by pitch (via elevator)
3) steering via bank angle, with automatic "up" elevator by virtue of altitude control
4) elevon and V-tail mixing is available in RC and autonomous modes
5) 50Hz attitude control, vibration tolerant thermopiles.
6) gain scheduling, 3 layers of PI and D.
7) Turn-rate wing levelling based on special data fit routines of the 5Hz GPS data for turn rate estimation.. it is not a MEMS yaw gyro like Pico N, however it has proven to prevent over-steering and is a key advance from circa March 2008.

It is a whole other universe. Don't read posts from before July of 2008, otherwise you will get errant information about Atto V1.7!!!

Fact is, AttoRTL hardware could be sold as a $100 programmable autopilot (17 grams including GPS!!!) that would do far more than the original Beta videos I posted in September 2007 (flying 15km over the desert). Back then no data logging, and gains were hardcoded into the code. Now you would have data logging, telemetry output, SET.txt access to plane setup parameters, and benefit of 14 months of Beta testing... truly a "proven" system and truly autonomous flight. With effective turn-rate data of my GPS plus analytical methods, no MEMs gyro is required (a $30-$50 component!!) so the cost savings is big. It does have its limitations because MEMS gyro is plane-centered reference, whereas GPS is Earth centered... and MEMS gyro can track rotation at 200Hz and my GPS is only good for this to around 0.5 second intervals... So if I do market Atto RTL as a programmable autopilot, you'll be constrained to stable motor gliders with positive pitch response to throttle, and flying in low wind.
Oct 17, 2008, 12:33 PM
Registered User
dmgoedde's Avatar
Thread OP

AttoRTL question


Someone asked me recently where Atto RTL is... basically it is on the back burner behind V1.7.

It could be made as a simple programmable autopilot (3D coordinates) for stable motor gliders... it would be the World smallest at 0.63 cu inch INCLUDING GPS and antenna and only 17 grams (2 gram PCB!!! PCB only 0.125 cubic inch, 1"x1"x0.125"), I might work on this a bit more and still get it out the door pre V1.7.

It has no thermopiles... so won't care about fog or mountains. It would have these features over and above another competing system:
Pluses:
1) Only ~ $100, but $200 TOPS. There is almost no hand assembly.. single PCB + automated pick and place, then adding bank of right angle header pins.
2) 3D waypoints (not simply lock at current altitude)
3) 5th decimal Lat and Lon (not 4th)
4) smaller (much)
5) data logging to removable SD (vs none), 30+ parameters (servo positions, all GPS data)
6) 5Hz attitude control (vs 3Hz)
7) at least 100 waypoints
8) user-configurability beyond simple trim pots... type in exact gain values to a simple .txt file on the SD card.
9) A true trigger of servo control or logic between WPs or at WPs. distance intervals, etc...... (not simply a line that goes high at a WP)


It would have only 2 drawbacks I can think of.
Minus:
1) turn rate is by GPS data fit to 5Hz heading data, not 200Hz (or whatever) integration of MEMS gyro
2) no airspeed control option, but throttle lock has proven effective. It is "proven".

Same:
1) Only for stable motor gliders


I have a lot on my plate... sometimes I need to get myself pumped up by writing stuff like this. If this envigorates discussion, then by all means great!

I would like to hear what people think of such a cheap autopilot. I flew a few hundred flights in 2007 (plus beta testers) with a system even simpler.

I do have several panels of PCBs for this on-hand, plus all required components. I also have the stencils ready. The only thing different between an AttoRTL and this is the code... proven code based on hundred of hours of flight, and hundreds of missions by several guys.


Quick Reply
Message:

Thread Tools