Discussion FLEXIPILOT 6DOF IMU autopilot and EasyUAV

[kbosak]

2 flexipilots stacked on each other in order to give the idea of compactness:




EasyUAV:


EasyUAV:


Google Earth data created automatically from flight logs:


I am releasing a few data about the autopilot you are using in EasyUAV.
Along with a few documents.

The initial idea was to make available an integrated solution.
You can consider EasyUAV a good starting point for UAV world,
either for amateur use of business evaluation of Flexipilot capabilities.

EasyUAV is shipping!

EasyUAV is sold assembled, tuned and flight tested
(incl motor, servos, ESC, installed autopilot, wings not glued).
The user needs RC TX, 6ch RX (PCM or anything with failsafe is a must for UAV),
Lipo3S around 2000mAh, own video system or camera.
450g of payload or ballast INCLUDING the camera.
You can connect to 2 servo triggers dangling in the nose section.
The autopilot is placed under the wing (the safest place).
For advanced users, waypoint scripting is possible using USB console.
Log download and parsing software (export to GoogleEarth) is available.
Made in EU, worldwide availability.
Price 1199EUR + taxes,
3-4 weeks leadtime (winter weather is the limitation, the assembled UAVs are not overstocked yet).

The front bay is left entirely free for the user:
you can put there your lipos, RVOSD, digital camera etc.
It has 2 independent trigger channels, logging,
USB connectivity with console.

You can watch the official movies:

Site photo mission

High altitude, pinpoint autonomous landing

Long range:

Repetitive

Flying a 3D pattern

Bad weather

EasyUAV uses 6DOF IMU autopilot and 5Hz GPS plus barometric altitude sensor.
This is a genuine design, not sharing any software nor hardware with anything else.
Not available elsewhwere.


EasyUAV manual and selected autopilot documentation:
easyuav-manual
flexipilot-advantage
flexipilot-features
flexipilot-installation
flexipilot-navigation
flexipilot-software
flexipilot-triggers
flexipilot-weather

It can land autonomously following full landing approach,
but due to wing loading and weak static thrust
of the small pusher propellers, the takeoff must be manual.

FLEXIPILOT has the following capabilities distinguishing it from similar designs:
-support minimalist airframes, maximize payload and range/duration (rudder and elevator planes at the moment)
-fully automatic initialization and calibration
-focus on gathering aerial photography data, triggers and automatic payload management
-minimal cross-section, a single, sturdy 50x120mm PCB board
-USB connection and real-time mission simulator
-designed for use WITHOUT ground station, using variety of relative waypoints
-fly many times from different locations with RC transmitter, a bag of batteries and the UAV itself


NOTE: FLEXIPILOT can control neutrally stable planes, but I WON'T ship the specific version nor support using the autopilot on them, because the tuning effort (and finally the support cost) is prohibitive. My focus is to support amateurs getting airborne fast, getting pleasure from flying and sometimes developing their skills/platforms necessary for light-weight aerial photography business.


Questions: I prefer to use kbosak-AT-box43.pl because of better traceability of communication, if the question is not of general nature.


Don't be scared of the not-working website,
there is an important integration project coming along which postponed the website. The product is mature and flew many photo platforms since 1.5 years in various weather conditions, has been introduced to the public after serious testing. The ground processing software is under development and will follow user's demand, based on experience gained using EasyUAV.

[docphi]

Can we order just the Flexipilot?



Found the price list...

[tekrunner]

Congratulations, you finally created your own thread regarding your autopilot. I hope your products quality doesn't match your etiquette.

[rbeall]

Ditto, good luck selling your product. I'm sure customer service will be interesting....

[jglenn]

Bosak:

Interesting system/autopilot, kind of pricey, have you considered a USA distributor? How much is just the board? Just a couple questions. I thought a perfect IMU could be put in a baseball, thrown spinning, and after landing would still know which way is up?

If that is true, why is stabilizing a neutral stability plane so hard? Does it have
trouble knowing when it is level, or how to get there, if upset? Is there drift
of the gyros that cause this? Cannot the accels at least give an indication when close to level, if not under centripital forces? If you are flying level, I would think the accels could tell you that, but when turning or diving/climbing they cannot.

Yes it is hard to put a mag sensor on an electric plane. Instead of the complex Honeywell serial comm types, I have experimented with the GMR class sensor chips for one axis, simple analog output. You can place the sense path either vertical or horizontal. The vertical use generally does not see the mag field from the power wires, if located away from them, but only tells you when you are not level, and by how much, but not how to correct it. This is the basis for magnetic rocket tip over detectors that deploy an apogee parachute, it works quite well.

Horizontal mode is more fun, you can use to correct the pitch, either up or down, to stabilize the plane. I did this on a hand launch glider last winter, and it worked well. At least in one direction. As you rotate around the compass dial, the signal changes, which sort of wrecks the solution. And as you say, adding the Emotor does interfere with it.

One solution might be to shut the power system off for a brief interval, say 4 times a second, and take the readings then. A mag sensor can also give you heading information. The expensive mil AP use magnetometers, do you think they are doing complex software corrections?

[kbosak]

Interesting system/autopilot, kind of pricey

Well, I disagree.
This is the Western World's least expensive full inertial autopilot right now and for some time to come. The turnover is expected in years.

This is also the least expensive autonomous plane capable of making geopositioned photos,
flying out of the box. The nearest competitor in payload and mission is cropcam.

The FLEXIPILOT board alone is 999EUR.

The IMU system is driftless, uses closely coupled GPS course and IMU,
uses Kalman for gyro drift elimination and has sutomatic temperature compansation.
The IMU is typically converged around 1-2s after power-up if the temperature didn't changed much (less than 10C).
Flight ready LED informs about IMU ready to takeoff.

The reason of not supporting astable plane is:
-people that fly strange planes make photos of inferior quality. this is contrasting with the main purpose of the autopilot.
-astable planes require several months of testing and tuning. same ppl that underestimate the price of autopilots, underestimate the time needed for autopilot integration, which count in months. I don't want either disillusioned clients nor those claiming success and showing seriously mis-tuned results.
-the support for astable planes should include finely-tuned 'direct pitch drive', this is not about the sensor, this is about how many feedback loops you use.

I thought a perfect IMU could be put in a baseball, thrown spinning, and after landing would still know which way is up?

In my opinion the perfect IMU is the one that coupled with feedback control algorithm used, provides widest convergence zone, while keeping accuracy acceptable most of the time.
Repeatability and precision of the sensor is not the same as stable behaviour of sensor+feedback control system, and the limitation is not in the control algorithm alone.
My main concern is reliability of convergence in standard flying conditions, which inevitably has mathematic limitations during knife edge flight etc.
An IMU that is working well standalone, can have artifact that add up when coupled with control algorithm. Demonstration of 'perfectly working IMU' is not a guarantee of working control system.

Magnetometers:
In my opinion because of bank angle, only 3D magnetometer is helpful,
additional mathematic hacks based on curretn location and a priori formulas that change every year
is possible but are not really worth investigating
because for any magnetometer you must know the angle first, using the IMU.
Once you have this math machinery running, using 1D magnetometer is not serious.
One day I will introduce a version with magnetometers, not at this price range,
but most likely it will fly on specifically designed airframe.

No US distributor is being considered simply because there is no point of giving a few % to the salesman if he cannot help the user support, unless we double-triple the price giving it back where it normally belongs.

[kbosak]

Originally Posted by tekrunner

Congratulations, you finally created your own thread regarding your autopilot. I hope your products quality doesn't match your etiquette.

tek, could you introduce yourself?

[kbosak]

Originally Posted by docphi

Can we order just the Flexipilot?
Found the price list...

You can order any time.
Flexipilot is ready for shipment.
Please contact me using kbosak at box43.pl
What kind of platform you intend to use, what is the 'mission'?
Normally I would recommend EasyUAV because:
for 200EUR more, you have FLEXIPILOT, servos, motor and the plane that is tuned and flies.
There is always some learning curve about the product. I think the fastest integration path is to examine working system, then remove FLEXIPILOT and follow my instructions and support on what to adjust in order to make it work in a new plane. The fuselage halves of EasyStar that is the basis for the EasyUAV are not glued, you can take electronics out at any time.
Basically I am adding equipped EasyStar at costs, the reasoning is if I invest some time in assembling it and tuning, I spare much more time in trying to give the basic understanding of the functionality to the user.

[jglenn]

Overall in the world it is competitive, I am just a cheap hobbyist.

So it takes a lot of work to tune it to fly a flying wing? Those are very popular,
it might be good to eventually have an auto-tuning routine, if that is possible.

Maybe have a temporary thermopile horizon sensor to give it more references,
then after it "learns", remove it. The Atto I thought could be used on any plane, but I don't know about his new IMU one.

The word "artifact" describing cumulative errors of the IMU/software combination is interesting. I last ran into that word in medical imaging work.

Good luck with your product. If you need a service/repair depot here in the US,
I do fairly intensive smd rework, of rocket altimeters. Sometimes just get a bag of chunks of circuit board..

[kbosak]

Originally Posted by jglenn

So it takes a lot of work to tune it to fly a flying wing? Those are very popular,
it might be good to eventually have an auto-tuning routine, if that is possible.

The only self-learning autopilots I know of that learn in flight are not sold in quantities, those are basically research units (saw one for quadcopters BEING DEVELOPED in Zurich, if they continues as they are ...means very fast progress indeed.. they should be available... in a decade.). No single operational military UAV is working like this, they are trimming out manually etc.

The problem with flying wings is that they can have momentums from propeller in all directions, and badly designed, can take truly several months to tune with IMU because they can have almost no damping. You can control astable plane but you need to know all frequencies and amplitudes for all actualtor axis, for all interesting flying states. Basically you need a lab for this.
Additional reason is very weak directional stability that can conteract adverse yaw which can resonate with feedback loop... All this giving a conclusion that
The solution is thermopile, but:
-only good weather
-not aerodynamically clean
-basically not for semi-pro use because of supposed reliability of the sensor
And the main thing is that when controlling pitch directly as the main drive (thermopiles), you never get optimal angle of attack. But you get the angle of attack good to maintain given, fixed airspeed and altitude.

Of course you can find the guys that did it with IMU and their flying wings. On well designed flying wing, after tweaking here or there, shouting loudly 'we did it' even if it is not flying calmly at all phases of flight, but who cares, achievement is achievement.

This autopilot has been designed for most efficient photoshooting, and the development continues. Supporting exotic platforms is not interesting for me unless I am being paid for my time. Flying wing are not really good for photoshooting because of limited payload volume, and because of very weak camera protection possibilities, besides parachute, which again is not a toy the amateurs really like to use. Considering aerodynamics theory, after consulting a fellow PhD in aerodynamics, there are cons and pros considering wind resistance for flying wings, and there are never-vanishing problems with directional stability - whatever you do, later or sooner you will be beaten by the lack of directional damping, the same feature that is used for assuming immunity from side wind.

Considering manufacturing, at quantities specific to autopilots, I will continue the assembly locally (shortest delays).

[workshop]

Basically I am adding equipped EasyStar at costs, the reasoning is if I invest some time in assembling it and tuning, I spare much more time in trying to give the basic understanding of the functionality to the user.

That is a brilliant strategy supported by a very reasonable assumption. There is other evidence of excellent business marketing skills in your announcement as well.

Many engineers, because of their perspective in the product cycle, believe that building it is the hard part and “the rest is easy”. “Build it and they will come” only works in Kevin Costner movies. I’m not saying that marketing is more difficult than engineering; it is however a complex discipline rooted in science.

Good luck, I wish you all the best.

Jeff

[jglenn]

Concentrating on photo planes is a good idea, that is a major market of UAV's.

I see your point on flying wings, I find them hard to pilot, being an old rudder/elevator guy. Sold the only one I ever bought, it could do 100MPH,
another problem for me. Speed sounded fun, but you need very fast response times in the pilot.

Could simulation help in tuning? Could RealFlight, or 3D Studio Max somehow interface with your software? Probably need a bunch of new code.

[kbosak]

Jeff,
thank you for encouragement.
I worked for some renowned telecom company and research and I know all too well that maintenance, non-uniform requirements are simply asking for troubles, both from the maker's point of view, but it also hits the client directly!
What I am observing is that in EU it takes one micropilot, one good PhD student and one semester in order to have UAV integrated, not necessarily flying wing. OK this goes on different pace but there are reasons - they are not just lazy kids playing with RC toys, they got unique chance to make an UAV and get a diploma. Yet, it takes months with existing systems, if you want to get the most optimal tuning!

When we talk about marketing, the situation is very difficult: because personal UAV is... let's call it... emerging... market, you cannot just find a reliable salesman, because there is nobody. There should be some kind of consultant group in China in order to make this profitable, mainly in order to screen the well-paid designers from the users (this is what big companies do, aren't they?). Unfortunately you cannot do it here as it is a kind of intensively-supported B2B (business2business) kind of product on multi-million budget that we move into low-cost, B2Consumer.
Net effect is that Consumers expect the same Space Shuttles to be available for them, copying the thinking of multiple groundstations, long range modems, antenna trackers... The list goes on.

[kbosak]

Originally Posted by jglenn

Could simulation help in tuning? Could RealFlight, or 3D Studio Max somehow interface with your software? Probably need a bunch of new code.

Well in order to support one new platform I should realistically make around 4-6 models with different motors and payloads, tune them all, observe the parameters (what changes most for given airframe class), make some manual tests with logging but if something bad emerges also some wind tunelling (incredibly expensive unless glued to national research project) at the end I should devise a specific set of PIDs that would work on variables that are invariant for those airplatform, and make self-tuning algorithm for the minor trimming of the remaining variables.
The controlled variables can by anything: from simple rudder deflection with some formulas relative to bank angle, airspeed, wing loading etc, to complicated asymmetric relative spoileron deflection ratio depending nonlinearly on arispeed and bank angle.
Using existing simulation software sounds hopeless, after consulting the researchers: no software supports reliably Reynolds number characteristic to our planes (too small), anyway typical research time based on theory is comparable to experimental testing, and at the end, the conclusion will be like:
'we support flying wings with dihedrals -0.1 to -4deg with elongations X to Y wing loanding N to M airfoils family Z, F, G'... an then the user asks: will my zagi fly well with it? The answer based on this data most probably will be: no, 95% of flying wings you see on the RC market have insufficient stability properties for reliable, efficient UAV flight.
Whatever I do, I don't want to approach the subject unless additional research opportunity arises.
Also the long-term amazement about flying wings that lead to no useful product is well known in aviation.
One thing for sure: flying wings like to fly fast on small range of angles, this makes them maybe easier for thermopile systems with direct pitch drive.

[clolson]

Originally Posted by kbosak

The problem with flying wings is that they can have momentums from propeller in all directions, and badly designed, can take truly several months to tune with IMU because they can have almost no damping. You can control astable plane but you need to know all frequencies and amplitudes for all actualtor axis, for all interesting flying states. Basically you need a lab for this.
Additional reason is very weak directional stability that can conteract adverse yaw which can resonate with feedback loop...

Having been through the design process for a 8' flying wing I will agree that everything you say is correct for a badly designed wing. But a well designed wing certainly is stable in pitch. A well designed wing can have excellent directional stability (just don't blindly copy a zagi and everyone else's flying wing.)

you never get optimal angle of attack. But you get the angle of attack good to maintain given, fixed airspeed and altitude.

Did you say what you meant to say in this previous sentence? I'm just an aerospace engineer wannabe, but for a given airspeed and altitude, an airplane is going to settle out at a particular angle of attack no matter what kind of controller is running (assuming the controller converges to the target speed and altitude values.)

In other words, (1) angle of attack, (2) speed, (3) rate of climb. Pick any two ... if you have already picked a target speed and a zero rate of climb (i.e. a fixed altitude) you don't have a lot of choice other than via a priori aircraft design with the angle of attack.

Of course you can find the guys that did it with IMU and their flying wings. On well designed flying wing, after tweaking here or there, shouting loudly 'we did it' even if it is not flying calmly at all phases of flight, but who cares, achievement is achievement.

I think you are making this assertion with just a little bit too much confidence. Certainly you can point to *many* mediocre examples that have been demonstrated here using both conventional airframes and flying wings. A well designed flying wing flies really well. Add a reasonably tuned autopilot to a well flying wing and you have a *very* nice autonomous platform.

Flying wing are not really good for photoshooting because of limited payload volume, and because of very weak camera protection possibilities, besides parachute, which again is not a toy the amateurs really like to use.

Tell me why a flying wing configuration has limited payload volume? Our flying wing has a blended body center section that can hold more volume than conventional aircraft I've seen with twice the wing area. Weak camera protection? Again, why is that?

Flying wings tend to be pitchier than conventional airframes because of their very short length, that's a reason they might not be as good for a camera platform. (See, I'm trying to help you out here.) :-) But that's the only reason I can come up with.

Considering aerodynamics theory, after consulting a fellow PhD in aerodynamics, there are cons and pros considering wind resistance for flying wings

Are you implying that flying wings have inherently more drag than conventional airframes? Who's aerodynamic theory are you quoting here?

and there are never-vanishing problems with directional stability - whatever you do, later or sooner you will be beaten by the lack of directional damping, the same feature that is used for assuming immunity from side wind.

You must be referring to a zagi or zagi like design here? A zagi is an example of a flying wing, but it is only one example. You can't extrapolate from there to all flying wings, just like I can't pick an example of a poor flying conventional airframe and extrapolate from there.

Regards,

Curt.