Autopilot Control Response

[flyzguy]

I am an aerospace engineering student at the University of Colorado building a UAV for a year-long senior project. www.peregrinereturnvehicle.org

Our mission involves detaching an autonamous glider from a weather balloon at the edge of space (92,000 ft) and flying at least 40 miles to a predetermined landing site. This involves flying at a wide range of reynolds numbers (10,000 to 2,000,000) and Mach numbers (0 to 0.45). Because of this wide flight regime, our Professional Advisory Board here at CU has filed an official request for action. We must either prove to them that our autopilot (micropilot 2028g probably) is capable of controlling our vehicle during every phase of the flight, or de-scope our project to only include flight-proven controllability.

The fear is that the gains we tune during test flights near the ground will not be effective throughout the entire flight. In other words, our autopilot will fly our glider well at the altitude at which we test it, but come 90,000 ft, it will either under or over-control our glider, resulting in our vehicle going way off course, or enetering a structure-risky spin or dive. (flutter is another concern).

It is my belief that what really determines the effectiveness of the control gains is the dynamic pressure on the control surfaces (elevons/rudder). Because a glider will naturally fly faster where the air is thinner, the dynamic pressure the vehicle sees should stay relatively steady, and the control throws should fly the plane well at any altitude. This can be shown theoretically that for steady glide through varying air densities, the dynamic pressure stays constant. Is this reasoning correct? I know that many control applications involve gain scheduling. Do we need gain scheduling?

Thanks for any input!

[GeorgeDog]

I think you must know about this guy who has done your proposed project several times over, but just in case:
High Altitude Glider Project (http://members.shaw.ca/sonde/index.htm)

It sounds like your review board is risk adverse and I think you need to accept that things are likely to go wrong on any project spanning such large ranges in operating conditions. I'd suggest that instead of trying to convince that board that your project will work as planned that you show how to manage failures, anomalies, and operational deviations should (when) they arise. Perhaps you could plan the flight out in the desert that is well within a weekend drive from your school. Seeking permission to fly at a government test range wouldn't hurt your post-college networking either.
This is an ambitious project to complete in a few months but very cool. Good luck and enjoy. -GD

(BTW, I think you are correct that control gains can be static but contact the guy above who has done huge amounts of research in this area).

[flyzguy]

Art's site has been a huge resource for us, but for some reason I never thought to e-mail him about the control gains.

You have a good intuition of our advisory board. Sometimes it can seem unfair when a PAB member asks for an analysis of something without first making an argument for why it is a risk. In other words, a PAB member could ask us to do a lengthy analysis on the chances of our glider being snatched up by aliens half-way through the flight. I think flutter and gain scheduling are worth looking into, but we have received other requests that have been a waste of time (skin friction melting our vehicle).

I like your idea of displaying our problem-solving abilities instead of trying to address everything beforehand. I'll try to post media next semester when we have something more interesting to talk about other than flutter...

Thanks

[Gary Warner]

Control surface effectiveness is proportional to indicated airspeed. Control indicated airspeed and surface effectiveness will be controlled. To control airspeed over 30,000 feet (where it looks like you want a 'free fall') use a speed brake design (split rudder, flaps, side fuselage panels, spoilers, etc.). Keeping the indicated airspeed close to the tuned airspeed for flight will reduce concerns of over control. Below 30,000 feet the navigation system can be used without drag components so that your L/D can be improved.

Flutter is not a big concern below 200 MPH with closed-tube stress skinned wings, balanced control surfaces and smaller AR wings. Still, proving at what airspeed a wing will flutter can be difficult. Even on full-scale planes the best the engineers can do is come close to predicting and usually over build to a higher than anticipated airspeed, based on know factors of proven flying surface designs. In any case, you will shoot for torsionaly stiff flying surfaces and control surface balancing.

It's too bad your 'customer' ruled out ballistic recovery. It has made you job much harder.

Gary
--

[Unterhausen]

I wrote a pitch flight control program for a simulator that had been called the National Aerospace Mole before I fixed it. It was a propulsion model of the National Aerospace Plane that had a rudimentary flight and atmospheric model. The problem they had was that once they got above a certain height, it would go unstable and their controller couldn't recover -- the plane would end up at negative altitude. I did end up scheduling the gains. I recall finding information that led me to pick a particular altitude to swap gains. Since your plane is subsonic and falling, you probably can live with much lower gains at high altitudes. I'm assuming it's going to be stable.

The safety weenies can be such a pain. I had a professor on my committee that liked to play a game where he would predict death and doom. When it didn't happen, he would insist that I got lucky.

[kd7ost]

I don't think gains minimized at higher altitudes is needed.

You will have a greater speed up high at terminal velocity. Maybe as high as 300 mph or above depending on pitch angle. I think the forces on the control surfaces at that altitude are the same as forces at terminal velocity at 1000 feet. Drag to weight, (With gravity supplying the thrust) is equal at terminal velocity in a given airframe irrespective of speed. Low atmospheric density at high altitudes means lower drag and lower forces which gives higher speeds till the drag equals the force of motion. Of course terminal velocity at 1000 feet will be much slower.

I think there are a lot easier ways to pull it off than a micropilot 2028 though. What control surfaces are you going to employ?

Dan

[flyzguy]

Well I e-mailed Art about our gain problem but I know he is a very busy guy these days.

The speed brake/split aileron/rudder idea is something I pushed for that ultimately got vetoed by the rest of my group for the sake of KISS. Right now, we have planned on elevons and winglet rudders. I think we should be able to deploy split rudder to slow down. This could be used for added stability and damping in recovering from a tumble in zero atmosphere after drop (92-100k ft). It could also be used to control our airspeed for flutter prevention and even control gain help as Gary suggested.

One of our faculty who builds UAV's at CU also commented that the micropilot is overscoped for our project. I'm probably the wrong guy to talk to about it though as I am more of a fabrication/aero guy and our systems engineer made the autopilot decision. I trust his judgement though I'm sure we could do our mission with less.

I am going to try to kill two birds with one stone this wekend by learning to cut foam with a hot wire while prototyping a 1/2 scale model of our bird. I don't have a digital camera but I'll try to snag some media if all goes well.

[kd7ost]

The temperature at that altitude is typically -60F or lower. Have you guys worked out your power system? Some batteries don't work very well when it's freezing.

What will you use to track the glider for recovery?

In my local area there is an Amateur radio group called the "TVNSP" or Treasure Valley Near Space Program. A couple of years ago I cut away a capsule via DTMF using a cutaway unit I designed. My capsule dropped at over 360 mph to start but slowed to 110 mph at 20,000 feet where I had the capsule deploy a chute for landing. Here's our web site. http://www.tvnsp.org/

The Photo gallery has a lot of pictures showing what we've done.

Dan

[Gary Warner]

Flyzguy,

(Don’t take any of this too seriously - I'm just 'fun' thinking)

About the tumble thing... with the atmosphere being 1/100th that of sea level at 100K feet mass inertia becomes a major factor in stability (or lack there of) during the initial acceleration phase of the free fall. Keeping moment arms of mass small will allow for a faster attitude correction as the craft gains aerodynamic effectiveness as speed increases. You also need large flying surfaces for pitch and yaw located at far rearward moments to gain stability as soon as possible. The proposed flying wing platform will not have low mass moment arms due to the aspect ratio of the wings. Also the design lends little to yaw stability at high altitudes. What were the main factors in deciding to use a flying wing platform?

It's hard to make assumptions that sound logical since I don't know the fine details of the project, but let me make an assumption, never the less. If there is a vertical free fall to 30,000 feet with no attempt to navigate, then at 30,000 feet navigation begins, you need only have a glide ratio of 7:1 to cover a distance of 40 miles. This kind of glide ratio doesn't require anything close to a sailplane design with a large aspect ratio. Since it looks like you are planning to use a "main chute" for the last several hundred feet for final recovery, flight to landings and slow flight speeds are not needed. Seems to me that a flight in the 500k to 1M Reynolds numbers is reasonable, with a final near-earth airspeed under 100MPH. This may sound fast, but if you deploy a two stage chute, deceleration could be well within limits.

Question about your navigation model. Are you planning to use the 'lawn-dart' navigation model (like the space shuttle)? I know you have a requirement to navigate up to 40 miles, but it won't always be needed. Obviously a horizontal navigation of say 5 miles will present a totally different flight environment then say the 40 mile horizontal flight. If using a controlled decent navigation model (like full scale gliders use) then airspeed is managed and a navigation model might look like a circle approach.

I can see where both models could be used to enhance accuracy. The lawn dart model can minimize drift due to prevailing winds and the circle to landing can improve control of airspeed. Are you planning to use both models and if so, how are you going to transition and at what height?

Gary
--

[GeorgeDog]

[[ (skin friction melting our vehicle) ]]

That is very funny :D I'd be more concerned about it flaking away as frost.

[[ If there is a vertical free fall to 30,000 feet with no attempt to navigate... ]]

He's building a glider, not a bowling ball. Using the thin air at altitude the vehicle will fly very fast to generate lift (and drag) and be able to cover huge distances before getting down to 30k.

[[ The temperature at that altitude is typically -60F or lower. Have you guys worked out your power system? Some batteries don't work very well when it's freezing. ]]

This is a very important point. Having experience in space systems design and operation I can categorically state that most of the problems arise when the systems are put into the thermal chamber and frozen for the first time. I'm sure you'll insulate the internal stuff but it doesn't take much of a thermal short to wick the little bit of warmth away. And if 90k in altitude is much like orbit the side facing space will freeze while that facing the sun will cook (can anyone vouch for this? There might be enough cold air to keep the illuminated side cool as well).

-GD

[kd7ost]

[

This is a very important point. Having experience in space systems design and operation I can categorically state that most of the problems arise when the systems are put into the thermal chamber and frozen for the first time. I'm sure you'll insulate the internal stuff but it doesn't take much of a thermal short to wick the little bit of warmth away. And if 90k in altitude is much like orbit the side facing space will freeze while that facing the sun will cook (can anyone vouch for this? There might be enough cold air to keep the illuminated side cool as well).

-GD

We've flown thermocrons to in excess of 100,000 feet. We've never observed any thermal temperature variations due to solar sources. We sample ambient air, and air temperatures inside capsules about 1 time per second.

I've purposefully sent up capsules in a parasitic fashion with open structures and no insulation to test batteries in the extreme cold. Batteries that don't work well at all are any standard alkaline or carbon zinc cells as well as Ni Cads. Batteries that do well are NiMh and LiIon's. The capsules like this aren’t “cold soaked” though. At ground level they leave a vehicle at the launch point at 70 degrees or so. It takes approximately 1.5 hours to reach apogee. Half that to get back to terra firma. So the cells and capsule parts aren’t in the extreme cold long enough to become frozen solid.

We have routinely operated servos above 100,000 feet so you should be good there.

We routinely observed that the coldest temperature in flight is right after the balloon bursts, or the capsules are cut away and the descent begins. We've argued as to whether this is because of fast air flow evacuating any heat or because the denser air of lower altitude is rushing into the compartment.

The ascent is very linear. We fill for 1000 to 1200 fpm ascent. It stays right in there till burst. The descent is very exponential. It comes down incredibly fast, even with chute deployed at first. As we get down to 35,000 feet we see it starting to drift down range more and more as the chute becomes effective in lower denser air.

Some GPS units won’t function above certain altitudes or in excess of certain speeds. Have a look at Ralph Wallio’s site for a list of known units. This isn’t a fully comprehensive list so you should plan several function test launches that don’t include flying the glider. You need to know if your electronics package can operate over the mission parameters as much as possible. Check on the left side of this page and select “GPS receivers vs 60K feet”.

http://showcase.netins.net/web/wallio/

You may have local balloon/HAM radio groups that launch in this manner. It would be a boon to hook up with them for experience. Maybe send Ralph an e-mail with your location and if anyone is there, he’ll know.

Make sure to review FAR101. That’s what you need to abide by to keep the FAA happy.

Dan

[Gary Warner]

He's building a glider, not a bowling ball.

Funny... :p

I get what you are saying, but this comes from his project documents on the listed web site.

I can see the bowling ball - bet that leaves a dent in the ground :eek: .

Gary
--

[GeorgeDog]

I get what you are saying, but this comes from his project documents on the listed web site.
--

I stand corrected. -GD

[flyzguy]

I am really enjoying all of the feedback on our project and this issue of controllability/flutter.

The flying wing was attractive to us due to mass limitationson our structure. The EOSS (edge of space sciences) balloon we use can carry 26 lbs. aloft, and after our science payload and all of our avionics, we are left with only about 7-10 pounds of structure to build our wing with. Eliminating a fuselage and tail also helped minimize ways that our glider could break. A conventional configuration would be easier to achieve large lateral stability margins, however. No freee lunch.

We've been working with the FAA because as you know, the FAA doesn't really know what to do about UAV's. We have submitted some documentation to them and hope to hear back soon. I'll feel alot better knowing we can actually do this.

Our avionics lead is overdesigning (IMO) some very heavy/complex heating systems for our bird. Hopefully, coupled with our foam construction, we won't have too many problems when we do cold chamber testing next semester. Right now, we are considering mounting our recovery chute inside of a canopy on top of the main wing. I am concerned that since it woll not be insulated, any moisture within the parachute fabric will solidify at altitude, and we could end up with a deployment problem when the time comes.

As far as tracking the glider for recovery goes, our customer from EOSS will be supplying a GPS beacon that is currently being used on their balloon flights. It currently ways 6 pounds with laptop batteries and an off-the-shelf GPS receiver, radio modem, and chute. We are hoping to strip it of weight as much as possible. Every ounce we save can be applied to making our plane more durable.

I'm not really sure what is meant by a "lawn dart" navigation mode, but it sounds something like an autonamous skydive. We considered making such a vehicle at first, with the mission goal of hitting any county road from 92,000 ft, instead of flying 40 miles to a road from 92,000 ft. I still like the solution in that it would require a minimum amount of airframe (building a steerable lawn dart). The downsides would be finding a way to control the vehicle and our customer would still have to drive a long ways to retrieve his science. In the end, our customer said "I don't want a lawn dart". I'm stoked anyway because I want to FLY, and not just drop.

I finished cutting foam yesterday and am looking forward to installing some spars and radio gear and going for some test glides!

[kd7ost]

You're in good hands with EOSS.

Dan

[Quacker]

Hi Flysguy,

Are you planning to downlink video?

[flyzguy]

For the time being, we will be recording video to a mini DVR with a 2 GB flash card on board. We can overlay GPS position as well as up to 17 other parameters with the micropilot GPS overlay board. We'll be able to view after recovery, but right now no downlink. I think our avionics team is looking into using the EOSS beacon to beam down video, and I really hope they figure it out. The hardware is there so I guess its only a matter of trying to code and decode the signal for the radio modem.

The protogrine (protoype peregrine) is done. It's a half-scale EPS glider we built this weekend. I was mostly happy with how it came out except for our rushed covering job. (VERY wrinkly). We went with ekonokote to save $$$ but for our complex shape I am starting to wonder if a more expensive covering material would have shruk up better. Pics to come soon. The glider seems to preform similar to most foamie wings, but we are still finding the C.G. optimum.

EOSS are good people.

[danstrider]

Just a few words of advice from someone who has experience with the Micropilot ...

Have you or someone on your team used it before? I have worked with one for a couple years to get it to the point of flying an airplane around in circles. Most of the time I spend was not tuning gains but rather just trying to get the doggone hardware to be happy. We had a lot of communication issues initally, for example not knowing the serial line needed to be a crossover configuration. My group got lucky with the gain part because we used a very standard trainer aircraft, but trying to get better performance was a really tricky thing even when the aircraft was flying well to begin with.

So I guess my sage advice is to make sure the Micropilot can boot up and work EVERY time on the bench and do exactly what you think it should do. Even then, in the air it will have different difficulties, so I reccomend using a standard RC system and flying the MP in datalogging mode to test comms. Then, once everything really really looks like it's working, put the MP in the loop and triple check everything including range checks and lifting the aircraft up and do a push test to see that the steering works (if you have landing gear). Then, go fly and have plenty of altitude when you turn it on.

For added safety during testing, I highly reccomend a way to override the Micropilot's outputs. Using RC per their advertised setup is somewhat scary being that if the MP goofs, you'll have loss of control. I have used the $100 insurance policy from this thread quite successfully in three different aircraft: http://www.rcgroups.com/forums/showthread.php?t=360746 I don't have ties to Jay, and I'm not advertising his product specifically, I just think prototypes need this type of manual override capability, especially for initial testing!

If you have any questions about the MP2028, feel free to ask!

Good luck!
Dan

PS: What DVR are you using to record and what camera are you using to provide the video for the DVR to record? The MP OSD works really well by the way :-)

[kd7ost]

Hey, are you guys still working on this project? Here's a blurb from some neighbors of yours that just lost a package. 350 fpm ascent speed is no good unless you're shooting for neutral bouancy or a long distance flight. That's way too slow. You need to shoot for about 800 to 1000 fpm and that will get you to 100,000 feet in about 1.5 hours. They said in the article that the descent was slower than expected. I'm sure they meant ascent. They should have had 1.5 pounds or so of free lift at launch time to keep the ascent rate up.

http://www.kwu.edu/publicrelations/stories/200607/balloon2.html

Dan

[Suresh]

llllll

[Suresh]

llllll

[Suresh]

Just a few words of advice from someone who has experience with the Micropilot ...

Have you or someone on your team used it before? I have worked with one for a couple years to get it to the point of flying an airplane around in circles. Most of the time I spend was not tuning gains but rather just trying to get the doggone hardware to be happy. We had a lot of communication issues initally, for example not knowing the serial line needed to be a crossover configuration. My group got lucky with the gain part because we used a very standard trainer aircraft, but trying to get better performance was a really tricky thing even when the aircraft was flying well to begin with.

So I guess my sage advice is to make sure the Micropilot can boot up and work EVERY time on the bench and do exactly what you think it should do. Even then, in the air it will have different difficulties, so I reccomend using a standard RC system and flying the MP in datalogging mode to test comms. Then, once everything really really looks like it's working, put the MP in the loop and triple check everything including range checks and lifting the aircraft up and do a push test to see that the steering works (if you have landing gear). Then, go fly and have plenty of altitude when you turn it on.

For added safety during testing, I highly reccomend a way to override the Micropilot's outputs. Using RC per their advertised setup is somewhat scary being that if the MP goofs, you'll have loss of control. I have used the $100 insurance policy from this thread quite successfully in three different aircraft: http://www.rcgroups.com/forums/showthread.php?t=360746 I don't have ties to Jay, and I'm not advertising his product specifically, I just think prototypes need this type of manual override capability, especially for initial testing!

If you have any questions about the MP2028, feel free to ask!

Good luck!
Dan

PS: What DVR are you using to record and what camera are you using to provide the video for the DVR to record? The MP OSD works really well by the way :-)



lllll

[danstrider]

Yes I have Suresh, but I'm a little bit creeped out that you're not saying what institution you're with or what your project is...

[Suresh]

Yes I have Suresh, but I'm a little bit creeped out that you're not saying what institution you're with or what your project is...

lllll

[danstrider]

Thanks for being a bit more open Suresh. This forum is definitely a place to be open rather than secretive.

Unfortunately I never played with the joystick interface to the Micropilot, so I can't help you there. Your best bet is to keep working with your friend to figure this one out; you probably can also get help from Micropilot if you call them. A word from the Micropilot experienced though, it does take a LOT of tinkering to get all the settings right.

Good luck!
Dan

[phubner]

A friend is ready to launch his second high altitude balloon project (see the first one at http://www.photoesque.com/HAPB ). According to current rules in the US, the payload can be less than 6 pound and there does not need to be any FAA paperwork. In fact, when he approached the FAA for flight registration and clearance, then specifically told him he should not file because of the weight rules. He actually got a few people that he was trying to file a courtesy plan.

Regarding Heating (or cooling!) he added a chemical heating pack (like a hand warmer) before launch. Thermal reading read high over the entire trip - but worse - caused significant condensation inside the cargo area due to the temp/humidity differences during descent.