Hi all! question regarding micropilot 1028

[uonuav]

Hi all,

I am a university student involved in developing a UAV.
There are some questions regarding the autopilot and cost that I would be interested in the opinions of you guys. :)
I have googled around for many various autopilot products and a problem I found is, I am confused with many autopilots products that the group control system cost almost as much, even more than the autopilot itself (such as the Piccolo system)

I think the Micropilot 1028 is the most suitable choice for our application, which is a light weight, cheap autopilot that can do some basics flying operations. It is to be equipped onto a model trainer. However, on top of the autopilot and the given software, what else is needed for such application? (I mean...for installing the autopilot, I guess there is an additional transmitter, receiver, batteries and cables, anything else?)

I have in fact e-mail micropilot asking for the details and pricing but I got no reply. I hope you guys can give me some advice. Any other cheaper autopilots?
Thanks in advance.

Stephen

[Unterhausen]

When I emailed them about the micropilot, they called me.

The problem I see with the Piccolo is the weight, but I think they may have addressed that.

The availablilty of software was why we went with the Xbow uNav. It's also fairly light, xbow flies theirs in a Zagi.

[uonuav]

o! do you mean this?
http://www.xbow.com/Products/productsdetails.aspx?sid=133

does it support video control and downlink or I need something separately?
:)

[clolson]

When I emailed them about the micropilot, they called me.

The problem I see with the Piccolo is the weight, but I think they may have addressed that.

The availablilty of software was why we went with the Xbow uNav. It's also fairly light, xbow flies theirs in a Zagi.

Have you had your MNAV up and flying yet? What airframe are you targetting? Have you done any autonomous work yet?

I too am very excited about the Xbow MNAV because of it's wide open software architecture, but we are currently being slowed down by some of it's weaknesses.

- problems with weak/insecure gps antenna connection
- problems with R/C receiver range once it's been wired to the MNAV
- insecurities about the built in manual override mode (especially in the face of poor R/C range issues.)
- questions about the quality of (or problems with) the ahrs/nav kahlman filtering code that computes a roll/pitch/yaw, lon/lat/alt estimate from the raw sensor data.

The MNAV has some real strengths and is a very cool device, but we are 1.5 years into it and still haven't flown autonomously because of some of these issues we are still struggling with.

The MNAV should be considered an autopilot "kit", not an autopilot. Some assembly required. Much coding required. A modertate amount of engineering required to make it fully operational and useful.

I'm still very hopeful, but a couple things have taken the wind out of my MNAV sails a bit recently. :-(

Curt.

[Unterhausen]

The micronav isn't easy, that's for sure. I was about to go into altitude hold mode on our first flight when I tipstalled the airplane. The winds were too high. Hopefully I'll do it today and give you a report.

The GPS antenna connection seems ok to me. A threaded connection such as an SMA might be a little better, but once you get the MMCX connector seated all the way, it's really hard to get it back out.

I just posted a longer answer about the receiver issue/manual/autonomous switching over on the Sourceforge forum. It does require some work, we're just going with the stock configuration for now.

I don't know what to say about the Kalman filter. It seems ok, but actually on our first flight the data didn't make any sense. Problems with the glow motor made it so we weren't paying enough attention to the computer, so I can't blame everything on the micronav. We had walked around a large circuit with the plane on a previous occasion, and everything looked ok. I don't know if you need to provide vibration isolation on a glow motor plane or not. Hopefully not.

People whine about the micropilot, but I don't know if that's due to brand loyalty by other vendor's customers or not.

[uonuav]

the micropilot seems to be one of the most economic choices, isn't it?

[lvspark]

UNAV is economical.

If I were to be shopping for an autopilot, micro-pilot would not even be on the list.

[uonuav]

oh! would you mind telling me why you wouldn't consider micropilot?
:)

[reedchristiansen]

Stephen,

You might want to take a look at Kestrel autpilot from Procerus. It is a bit smaller then the micropilot and has lots more features. It is similarly priced.

http://www.procerusuav.com/

Feel free to ask me questions about it, I work for the company.

Reed


Hi all,

I am a university student involved in developing a UAV.
There are some questions regarding the autopilot and cost that I would be interested in the opinions of you guys. :)
I have googled around for many various autopilot products and a problem I found is, I am confused with many autopilots products that the group control system cost almost as much, even more than the autopilot itself (such as the Piccolo system)

I think the Micropilot 1028 is the most suitable choice for our application, which is a light weight, cheap autopilot that can do some basics flying operations. It is to be equipped onto a model trainer. However, on top of the autopilot and the given software, what else is needed for such application? (I mean...for installing the autopilot, I guess there is an additional transmitter, receiver, batteries and cables, anything else?)

I have in fact e-mail micropilot asking for the details and pricing but I got no reply. I hope you guys can give me some advice. Any other cheaper autopilots?
Thanks in advance.

Stephen

[lvspark]

oh! would you mind telling me why you wouldn't consider micropilot?
:)


I have read many papers that report difficulties getting it to work. Could be that they give them out to many university teams and the teams can't figure it out, or could be the MP device itself? I don't know, but there are plenty of negitive reports out there if you go looking for them. It shouldn't take 10 people a year to get an auto pilot to fly a plane reliabily. (I do think it is great they support the university teams.)

I have also had a few people contact me to assist them with issues they had that were not resolved by MP or were not explained well in the manuals. (more evidence of people having difficulties with it)

Lisa Shaw of MP also contacted me and gave me a hard time for stating my opinion of MP on a public forum. She was not nice about it.

So, 3 reasons they would not be on my list.

I told them to send me 1 for evaluation and testing, and if it was a good unit, I would change my tune. Until then, the only data I have access to indicates its a POS.

[AntonK]

I worked with the MicroPilot 2028 and I can say it is a horribly designed piece of hardware. The EMI from the unit is so bad that the GPS on it cant even aquire. Just FYI the documentation for the 2028 states that on a cold boot, the GPS could take as long as 45 minutes to aquire a 3D lock. Their customer service is very good though. However it doesnt overcome the fact that for the price there are better products out there. The 28 grams is very false, once you add the wiring harness, GPS antenna, servo bus board your looking at at least 70 grams.

With no idea the size of the model your looking at using, its hard to say what would be best. It seems that the Piccolo LT is now available which is a pretty good deal for the size. I havent used the procerus but have seen the University of Florida use it in an MAV, and watched all the cool BYU videos :). As some of you know Im part of the development team with Paparazzi, so I have my own bias. Ill try to leave it out of the mix though, unless someone has questions.

As a plug though, The Paparazzi team did fly 5 aircraft at 1 time, through 1 groundstation during the EMAV06 competition earlier this month! The groundstation operator needed a cigarette after he finished!

Anton

[uonuav]

Stephen,

You might want to take a look at Kestrel autpilot from Procerus. It is a bit smaller then the micropilot and has lots more features. It is similarly priced.

http://www.procerusuav.com/

Feel free to ask me questions about it, I work for the company.

Reed


Thanks Reed
I have in fact been to the website b4 and found that it costs around $7000 to include the software. I think this is quite a bit higher than expected, although Procerus is much better in ability I think

[uonuav]

I got an reply from MP that they said they can offer discount up to 40% to student projects. This makes a lot of different to other autopilots (although I know Piccolo does it too, they are very expensive equipments)

[corosplat]

The EMI from the unit is so bad that the GPS on it cant even aquire. Just FYI the documentation for the 2028 states that on a cold boot, the GPS could take as long as 45 minutes to aquire a 3D lock.

with repsect, that's just nonsense. I guess you had other problems with your installation. With a cold start, meaning you unplugged the GPS backup battery, it should take at most 15 minutes (mp2028 manual p. 9), that's not a MicroPilot problem, it's just a Trimble OEM module. In normal operation it takes about 20 seconds to acquire lock.

Despite the bad rep on these forums, my experience with the mp2028 has been fairly good. Installation and configuration is straightforward if you follow the manual. I had no particular problems with interference that is often reported.

Tuning the various control loops is the big challenge. With about 12 different control loops, each with PID parameters, the possibility to schedule gains by speed, and other control tricks like feed-forward, anti-windup and output limiting, it is an enormous parameter space to try to optimize by hand. The ad-hoc approach to tuning suggested in the manual just doesn't cut it. You really need a more systematic approach and/or a good simulator. It doesn't surprise me that a lot of people get stuck at this stage, never get it working right, and blame it on the hardware.

My only complaint with the MicroPilot is that the GPS isn't tightly integrated with the inertial measurements. Positions (and heading) are taken as-is and not estimated from all sensor input. This means the heading control loop especially suffers from a slow update rate and GPS time lag, degrading turning performance.

Cheers,
Splat

[reedchristiansen]

Procerus has a fairly substantial university discount. Atleast %20 if I remember right.

FYI....

Good luck,
Reed


Thanks Reed
I have in fact been to the website b4 and found that it costs around $7000 to include the software. I think this is quite a bit higher than expected, although Procerus is much better in ability I think

[lvspark]

Thanks for the input corosplat. Been a while since I seen you around.. ;)

[Unterhausen]

Now I see why everyone has trouble with the micropilot: too much work. I know people have problems with ground robots even when the companies selling them hold their hand. The academic discount for the micropilot basically gives them the rights to your source code and your gain settings. So hopefully they are accumulating information about airframes as time goes on.

[danstrider]

Thought I'd chime in about the MP2028, since I've been flying one in a school project aircraft for three years...

Setup was the hardest part for me, not having much electronics experience at the time. I put the MP and servo board inside an aluminum enclosure more for handling and mounting purposes, but also never saw EMI as a problem. The hardest part I thought was getting communication between the ground station and the onboard wireless card to be reliable without crashing the MP. I remember I had a hard time getting through the full wizard setup for the MP without getting "Runtime Error 0" (to this day, I don't know what that was, but I remember it!). But now that it is set up, I haven't seen that error; granted, we're to a new MP code version anyhow.

In the air was a fun story. Took the MP along as a passenger for a few flights and tried to check the datalogger. Then finally put the MP into the loop and flew around keeping manual control. Ended up having two major radio hits and (fortunately) landed safely. Turns out the servo battery and the autopilot grounds weren't tied together. With that fixed, it worked.

We picked a standard trainer (the Senior Telemaster is my best friend!), so we had to do no PID tuning. The generic gains they include are for a 40-sized trainer.

Life was easy after this point. Fly circles, start working on guidance strategies and payload control.

About 8 months was spent trying to get the MP2028 to work in the airplane. Once the hardware was hooked up correctly and the thing booted reliably, time to autonomous flight was about two weeks.

...

So there is another first hand story about Micropilot you guys can all add to your 2-cent advice.

Personally I like the Piccolo better, having worked with that one for around 4 months. The generic Piccolo complaint is that the Operator Interface is tabbed and hard to work with quickly, but I'm sure Cloud Cap knows that and is thinking of changes for later down the road. I am happy with the Piccolo controller and the HWIL option really is a terriffic option to have. If you have a higher fidelity simulator than FlightGear, you can make the HWIL option even more useful! I've logged hundreds of hours on the sim and feel confident flight will be simply a real-life version of software I already know will work. HWIL setup is worth the investment.

...

So my experience is with MP2028 and the Piccolo, but I have witnessed several Kestrel and homebrew autopilots fly around. (I haven't seen the Paparazzi or Unav fly.) When working, they all seem to perform about the same, though the higher-end seem to have tighter control of the flight path and attitude. In general though, it seems to be just a matter of how much time vs money was invested. More initial investment, the less time needed. Less up-front investment, generally the more time needed.

You might also consider which options you need. For a turn-key solution of a standard aircraft, the Micropilot isn't a terrible choice. For much greater control over the flight controller, the Micropilot wouldn't be a good choice, maybe in favor of the Kestrel or Piccolo option. So coming up with a list of the options you need and the options the various autopilots all have may be a good way to go about this choice.

Way too long of a message... hope it helps though!

Dan

[Unterhausen]

Dan, thanks for the thoughtful post. In the end I didn't buy the micropilot or the piccolo because of management resistance. I thought there were enough people out there that had reported reasonable results with the micropilot to discount the griping. I've seen a few posts along the line of "we had a micropilot and it didn't work very well," but it's hard to know what that means. The Cloudcap seems to be universally admired, but I hate to have that much money flying around in an aircraft that I have anything to do with maintaining.

[danstrider]

I hate to have that much money flying around in an aircraft that I have anything to do with maintaining.
Hehehe, just part of the business :-)

[clolson]

Personally I like the Piccolo better, having worked with that one for around 4 months. I am happy with the Piccolo controller and the HWIL option really is a terriffic option to have. If you have a higher fidelity simulator than FlightGear, you can make the HWIL option even more useful! I've logged hundreds of hours on the sim and feel confident flight will be simply a real-life version of software I already know will work. HWIL setup is worth the investment.


Dan, just curious. Is the Piccolo HWIL software already setup to talk to FlightGear? For what it's worth, with FlightGear you can go in and build a dynamics model for your UAV with as much fidelity as you have data for ... you need some knowledge of FlightGear and a lot of knowledge about aerodynamics, but it is very doable. I've been developing a dynamics model of the Sig Rascal 110 which still needs some work, but I'm close to being able to collect the needed data to really dial in my dynamics model. Right now it "feels about right" and is in the ball park for the numbers I know or can estimate, so the next step is to tune it based on actual performance data.

Regards,

Curt.

[builder]

The MP1028 is a reduced feature version of the MP2028 but is physically identical. The feature reduction has been done in the software. The navigation and flight performance is similar. I have worked with the MP2028 for the past 2 years and find that anybody that badmouths it has either not worked with it or has simply given up on tuning it because they didn’t get perfect results a few days after they started flying it. The general consensus from users is that it is a little hard to tune, but fly’s excellent once you have it set up for your particular application.

Over the last year I have set up a dozen identical UAV’s with a 2028 and have had no issues with any of them. The initial cold start usually takes about five minutes, after that all of them had repeatable gps locks within 1 minute.
Other than the initial minor servo trim difference between each UAV, all of them had identical files loaded on the autopilot, and all of them flew EXACTLY the same. The repeatability was 100 percent. I have enough confidence in them that the last 6 of the UAV’s have had their very first flight done autonomous from a hand launch.

I usually wind up being a safety pilot when long flights are done in the field. After the launch, I turn off the RC safety transmitter and get comfortable in a lawn chair and do nothing except watch the UAV for the usual 25 minutes. The only time I take over is to land in tight spots or to avoid power lines or trees in the landing path.

Here is a screen shot from the data logger of a flight flown over a section of land for a photo survey. As you can see the performance is excellent. So perhaps the best thing I can say to sum it up is “Don’t knock it till you try it”.

[danstrider]

Dan, just curious. Is the Piccolo HWIL software already setup to talk to FlightGear? For what it's worth, with FlightGear you can go in and build a dynamics model for your UAV with as much fidelity as you have data for ... you need some knowledge of FlightGear and a lot of knowledge about aerodynamics, but it is very doable. I've been developing a dynamics model of the Sig Rascal 110 which still needs some work, but I'm close to being able to collect the needed data to really dial in my dynamics model. Right now it "feels about right" and is in the ball park for the numbers I know or can estimate, so the next step is to tune it based on actual performance data.

Yes, the Piccolo HWIL sim is already set up to talk with Flightgear. It takes a bit of reading CC's documentation and trying it a few times, but it is supported as supplied. You do have to spend a bit of time building an aero model on Piccolo's end, not sure about the FG modelling. CC supplies a batch file that I would assume loads the Piccolo aero model into FG. Though Flight Gear isn't what I would consider an extremely high fidelity sim, it does seem enough to get within the ballpark. I have been using a different sim, but I went through the setup with FG to understand what the process was all about.

You competing in a student competition by chance? I know a couple school groups using the large Rascal; curious if you're involved in one of them. I'm especially interested to find out who all is using a Piccolo and to what level ... I never could convience my old group to spend that much money on an autopilot, yet it seems it would have answered so many problems.

Did that answer your question btw?

Very neat image builder! Good to hear you're having such good luck with it. May I ask what vehicle you're using or for some pics of it?

Dan

[lvspark]

Very neat image builder! Good to hear you're having such good luck with it. May I ask what vehicle you're using or for some pics of it?
Dan

www.cropcam.com

[builder]

Hello Dan, the airplane is an off the shelf Electra pro with a few mods for the autopilot and camera gear. In other words, a CropCam. You will notice in the pictures the camera (S5I) is mounted under the wing in a box to protect it during landings. There is no room in the fuselage, the other components take up all usable space. It doesn't seem to hurt navigation, and it's easy to access after the landing. The picture of farmland is from the same flight as the datalog. There are 48 pictures stitched together to make up this full section photo. Low resolution here of course, the full size file is huge as each seperate image is about 2 MB. With light winds the setup will fly the section in about 25 minutes and use up slightly less than half the available battery capacity. The climb to 2100 feet is the big energy user, once at altitude you could fly two sections back to back without a problem.
At 33 knots cruise you can cover a lot of ground before the batteries are drained, and it glides a long way when you start at that height.

[clolson]

Yes, the Piccolo HWIL sim is already set up to talk with Flightgear. It takes a bit of reading CC's documentation and trying it a few times, but it is supported as supplied. You do have to spend a bit of time building an aero model on Piccolo's end, not sure about the FG modelling. CC supplies a batch file that I would assume loads the Piccolo aero model into FG. Though Flight Gear isn't what I would consider an extremely high fidelity sim, it does seem enough to get within the ballpark. I have been using a different sim, but I went through the setup with FG to understand what the process was all about.

You competing in a student competition by chance? I know a couple school groups using the large Rascal; curious if you're involved in one of them. I'm especially interested to find out who all is using a Piccolo and to what level ... I never could convience my old group to spend that much money on an autopilot, yet it seems it would have answered so many problems.

Did that answer your question btw?


For what it's worth, if you dig into the FlightGear flight dynamics configurations, you have the ability to setup extremely high fidelity models ... as good as your data. This takes effort, but the underlying code is there to allow you to model as much detail and fidelity as you have data for.

Other flight sims "claim" high fidelity, but you need to know what you are getting there ... If you are referring to X-Plane, then know that it makes a lot of assumptions that are pretty good, but that only takes you so far. If you need to fix problems or differences in what X-Plane produces compared to what your real airplane does, you need to dig in and make black magic changes to your aircraft to have the results you want/need.

With FlightGear we use standard aerospace engineering approaches to define the performance and behavior of a model. That takes a bit more knowledge and work, but gives you the opportunity to do a *much* more accurate job if you want to dig in and do it.

For what it's worth, I visited lockheed-martin a year or two ago and got to fly a demo of their UAV HWIL system using X-Plane ... neat system modeling a 55lbs UAV ... I tried to land on a 5000' runway, was doing about 45 kts and 10' AGL when I crossed the threshold @ idle throttle. Halfway down the runway (2500' into this adventure) it became clear I wasn't going to lose enough altitude/speed to land (amidst the jeers from my buddies in the background.) Again this is idle throttle. So I pushed the nose down and tried to plant it on it's gear before I ran out of pavement, but bounced back in the air and by the time I crossed the end of the runway I was 50' in the air and 10 kts faster than when I started ... again at idle throttle ... and Lockheed-Martin was really gushing about how great their system and X-Plane's flight dynamics were ... and clearly I have no clue how to fly an airplane ... but I do know a little bit about energy and somehow they managed to produce a simulation of a perpetual motion machine so that was cool. I somehow suspect that the real thing would bleed off airspeed just a bit faster than the sim though...

So when people talk about flight dynamics fidelity, that really needs to be qualified. With any system, the result is only as good as the data you put into it. And with any system, you need to explicitely measure the "fidelity" against the real thing before you can make any true determination of faithfully the simuation represents the real aircraft in the real world.

I've taken FlightGear through the process of getting a set of cessna 172 flight dynamics FAA Level 3 FTD certified. That involves matching the performance of the sim to the performance of the real aircraft across about 30-40 different tests which have to match the real thing within some tolerance. (And then doing all the paper work up/down/left/right to support your results.) FlightGear will take you as far as you want to go, but you may have to put in some effort to get there. X-Plane involves a lot less effort, but I'm a little doubtfull as to exactly where it takes you and how far ... and you may want to have your chicken bones handy.

Personally, my observation is that X-Plane usually comes up with a pretty good guess out of the gate, but you have to do a lot of voodoo to dial in the performance to closely match the real thing. And if you don't do the voodoo, you may end up with some weird things, especially with smaller scale aircraft. And if your simuation doesn't match the real world within reasonable tolerances, then any gain tuning and autopilot development you do in the lab is going to have to be refined substantially once you get out in the field.

With FlightGear, it may be more work to achieve a high level of flight dynamics fidelity, but you are using standard/accepted aerospace engineering practices to get there.

For what it's worth #1, I'm full time staff (Mechanical Engineering) at the University of Minnesota. I have a fair amount of R/C experinece so at times I've been paid a small portion of my salary to help the Aerospace Eng. department with one of their UAV projects ... they have 2 rascal 110's they are flying. Flying is often where student led projects fall on their face, so having a full time staff make that part happen has really helped this project be a lot more successful than it otherwise probably would have been. :-)

For what it's worth #2: I know a guy who is selling 4 rascal 110's and would probably offer someone a pretty attractive deal if they bought the whole package. He is moving and needs to sell them. Let me know if anyone out there is interested, I can connect you up.

For what it's worth #3: I apologize for the overblown flight dynamics fidelity rant ... but I've got to stick up for my baby. :-)

Best regards,

Curt.

[danstrider]

Thanks for the information builder! I read through the website too and feel like I have a much better idea what you're doing now. Neat that you're doing production with the MP.

I have a question how you do the image mosaics. I read "The raw images acquired by CropCam are not orthophotos. Orthophotos are the result of the orthorectification process which involves standardizing photo scale across a photo which could have varied in the raw imagery due to variation in the topography of the terrain being acquired. Imagery from CropCam could be orthorecitified if a digital elevation model (DEM) of the area existed or was collected."
I understand the variation of aircraft attitude for a fixed mount not giving exact ortho-images. How do you correct for this in the mosaics? Do you compare the datalog to the time stamp of the photos and use some post-processing software to orthorectify the images? Or is that done by hand? Do you have access to the algorithms that do the transformation on each image to create a mosaic? The mosaiced images look awesome I must say!

The reason I'm curious is that I've been involved with a student group working on real-time image mosaicing using nearly the same setup you're using. The one idea we came up with that you might find useful is using a tilt-tilt mount for the camera and using a table lookup to convert roll and pitch directly into servo positions (i.e. offset angles) so all the images are orthogonal wrt earth. We have a real-time downlink for the imagery and a custom viewer GUI, but the piece that is still lacking is the real-time mosaicing software. I am not as involved with the club as in the past, but you can see some of the systems I helped develop in the gallery, especially the "Nikon D50" and "Ortho Optio" folders: http://www.ncsu.edu/stud_orgs/ar


Thanks a bunch for the extensive post Curt! It is nice to know Flightgear has more potential than I first gave it credit for. I am actually not using X-Plane but a NASA Dryden sim. But I have taken the X-Plane and Flight Gear info to heart. I had no intentions of running FG into the ground; for my purposes the other sim was already set up, so my impression was biased since one worked and one I had to make work. And FG seemed to work okay, so I have no complaints.


Sorry to uonuav for the off-MP discussion. Glad to be a part of accumulating this level of knowledge on RCG though! I have learned stuff already :-)

Dan

[clolson]

Talking about mosaicing ... I've got an idea that's not fully thought through, and I haven't seen how the really smart people do this but here is what I've been thinking about.

FlightGear has the ability to return the lon/lat/elev of any terrain point you click on. Essentially it projects a vector from they eyepoint through the x,y screen space where you did your mouse click, into the 3d world and finds the intersection point with the terrain. (And note that the FlightGear world is derived from the SRTM2 data set so it is pretty accurate.)

We have demonstrated that we can take our IMU/GPS data and render a synthetic view that is *very* close to identical to what the live camera is seeing. Scroll down to the October 26, 2005 entry here for example (note that we haven't done a lot of fine tuning of the calibration yet):

http://baron.flightgear.org/~curt/UAV/Rascal110_2/

If we could figure out the exact screen space x,y coordinates of the 4 corners of the video image such that the video aligned as close as possible with the flightgear synthetic view, then for each image, we could project vectors through those 4 corners (just like doing 4 mouse clicks at the corners) and get back the real world lon/lat/elev of the 4 corners of the image.

That ought to be enough (assuming no lens distortion effects) to position and rubber sheet each image so it is approximately orthorectified.

If terrain is a factor, you could start dividing the image up into a grid pattern and find the real lon/lat of each individual grid intersection point.

Sounds simple enough in theory and most of the hard stuff is done and working.

If you had a frame grabber and drew the live video as an inset in the flightgear synthetic view, then you could click on a point in the live video and get back it's lon/lat/elev. We have a student here who is working on adapting the tiger street database so that we can do real time address lookups given a lon/lat. So put it all together [hopefully someday] and we ought to be able to fly our uav around and get back the street address of any thing we click on in the live video. Should be cool when/if we pull it off. :-)

Regards,

Curt.

[builder]

Do you compare the datalog to the time stamp of the photos and use some post-processing software to orthorectify the images? Dan


Hello Dan, the image manipulation software is not my strong point, but I believe it is done as you mentioned. The GIS guys I have talked to have also used the method of getting gps coordinates from the corners of the section and rubber stamping the image with known GPS data.
The image stitching software is Autostitch by the way, we have tried PT-Gui and it works well but is more dependant on image overlap and seems to have trouble with more that 2 dozen photos. Autostitch is a real processor hog, it makes my laptop unusable for anything else until it is finished. Wish I had dual processors......
Interesting photo gallery. One thing I can say about this work is that there is always something new to learn. I will PM you with more info.

[Miguel.0]

Hi every one, i am new to the forum. I am working whit an MP2028 on a delta wing UAV, wing span 60 cm (24 in). We succesfully fly in autonomus mode, but we can't acheive the extremely good result shown on the datalog above (the Builder one). The longitudinal control is ok, we can mantain the altitude whitin 3 feet but the flight path is not very precise.
Builder can you send me the flight plan (.fly) you used in the flight reported above? You are using nothing more than fromto command?
Thank you

[iflytrikes]

Hi Stephen
Im on the verge of getting an MP2028 but came across your suggestion for the Kestrel ... could you give more details ....
Neville
see my blog
Stephen,

You might want to take a look at Kestrel autpilot from Procerus. It is a bit smaller then the micropilot and has lots more features. It is similarly priced.

http://www.procerusuav.com/

Feel free to ask me questions about it, I work for the company.

Reed