Newbie Needs Help

[NewBee123]

Hello,

Hello,

I am new to this forum but not new to RC flying. I work for an avionics company as electrical engineer. Lately I have developed interest in UAV for aerial photography. I did a few searches for info, but it did not turn out helpful. Can you tell me where to begin with? Any books/links or any other material that I can read to get started in the world of UAVs? Any flow chart that describes the process? Also, what kind of aerospace background I need to develop to get into this field? I will appreciate your help.

[JettPilot]

This forum has all the info needed :D If you have a question or run into a problem just ask and someone ususally helps out. Once you start understanding the electronics, its not that hard :p

[kd7ost]

Give this place a try as well. This is the forum for the Remote Control Arial Photography association. It runs the gamut of newcomers to experts. Some pilots on the RCAPA forum earn a living with it and there are many just in it for the hobby. Fixed wing and Helicopters are represented very well with a lot of professional advice available.

http://www.rcapa.net/forums/

Dan

[NewBee123]

Thank you guys, for replying back. The thing is, I have some background knowledge, but I just do not know where to start. How would you approach this if you were in my shoes? For electronics, what is the starting point? What should I buy first as the building block? I will appreciate ideas..

[kd7ost]

Thats not an easy answer as there are so many viable options. It would be better to start by defining what you want to do? Where do you fly from? Do you have any equipment to start with? Can you fly gas, glow or electric where you are? How much time do you have to spend on it? Some ways are quick and easy and some take more work. Can you transport a big plane and do you have room to fly it? Do you want to take pictures of houses, parks etc? Helicopters are also an option. I still think you'll get better AP advice either on the RCAPA link I posted above or at http://www.rcuniverse.com/forum/tt.asp?forumid=81

Guys here are the UAV crowd and even though some are very skilled at pictures, you will find a more diverse cross section of AP builders and flyers at those other two places.

Dan

[LukeZ]

NewBee, I agree with Dan. At the risk of sounding off-putting, if you don't know where to start yet then the place really to start is with reading - lots of reading. See what other people have done mostly, and that will give you an idea of where you'd like to go. Here's a few projects others are or have worked on, to give you a start:

Glidersonde: http://members.shaw.ca/sonde/index.htm

Chris Good's UAV: http://www.chrisgood.com/uav/index.html

Low Cost UAV: http://gs3636.sp.cs.cmu.edu/uav/index.html

Paparazzi: http://www.nongnu.org/paparazzi/

Most Evilist UAV: http://www.weedplane.com/uavproject/index.htm

I could list more, and none of those sites will have all the information you need, but it will give you a broader overview of the sorts of options and avenues you might take. After you have a general idea of that sort of thing, then you can start getting into more specifics.


Luke

[NewBee123]

Thank you Luke and Dan for your posts. I am reading the links that Luke provided and they are very helpful. I have a stupid question though: is there a good reference material available on the working of Gyros and Accelerometers? I still have not fully understood the working of these two things. I will appreciate your ides.

[LukeZ]

NewBee, at least for me these are the most difficult things to grasp so you're not alone. I can't even say I know a whole lot about them, as I have yet to get to the stage in my autopilot design where I am dealing with them.

My briefest explanation is that a gyro senses deviation from a position - so it would sense if a plane banked to the left or the right, and by how much.

An accelerometer measures, of course, acceleration. If you like you can think of an accelerometer as measuring the "rate" of deviation - how fast or slowly is the plane banking to the left or right, for example.

Accelerometers and gyros are often used in conjunction with each other in projects such as an autopilot. An accelerometer works much more quickly than a gyro, and can detect almost instantaneously very small movements. However, an accelerometer is subject to error, or drift, over time. Gyros, as I understand it, are used to help remember exactly what "level" is, and can be used to offset the errors inherent in cumulative accelerometer measurements.

Of course I could be way off here, but this is what I recall. Others can feel free to chime in and correct anything I said.

Luke



A little bit on gyros:
Analog Devices iMEMS Gyro Concept Info (http://www.analog.com/library/analogDialogue/archives/37-03/gyro.html)
Gyrobot (http://www.barello.net/Robots/gyrobot/index.htm)
nBot - a two wheel balancing robot (http://www.geology.smu.edu/~dpa-www/robo/nbot/)
RC Groups Discussion - Gyro links (http://www.rcgroups.com/forums/showthread.php?p=1849414&page=2&pp=15)

A little bit on accelerometers:
Analog Devices - What is an accelerometer and when do you use one? (http://www.analog.com/UploadedFiles/Technical_Articles/599120474accel.pdf)
Analog Devices - Accelerometer Design and Applications (http://www.analog.com/Analog_Root/static/library/techArticles/mems/sensor971/index.html)

[kd7ost]

There are two types of Gyro's. Rate and vertical. Rate gyros are the ones we use in the RC hobby. They resist motion from a current position. They accumulate error though and will just as easily prevent motion from a new position in the axis. Vertical gyro's can be used to measure roll, pitch etc and won't accumulate error but are quite cost prohibitive at our hobby level. We're talking thousands of dollars for a single gyro.

Guys are using MEMS accelerometer type sensors and fancy software called "Kalman filtering" to achieve Vertical gyro type results from accelerometer type devices. That stuff is over my head though. I'm going the ultra simple route of roll control with a co-pilot, altitude control with a barometric sensing unit and steering through a GPS device called a PDC-10. It's pretty entry level work.

Dan

[vpatron]

Luke, Dan, great info on gyros. Thanks!

-Vince

[Unterhausen]

the way I think of gyros is that they measure the rate of rotation. Accelerometers measure linear acceleration. The reason accelerometers drift so much more than gyros is that you have to integrate them twice to get position, and thus any bias gets added up twice, which will eventually cause large errors. The micromachined gyros that we can afford are rate gyros, i.e. they measure the rate of change of an angle. So to get an absolute angle, you only have to integrate once. Thus any bias error is only integrated once and not twice. A mechanical gyro actually spins a disk in a gimbal setup. Thus the disk resisted rotation and basically stayed at the same orientation. The gimbal joints would have encoders, and thus you could measure the angle you had rotated directly, no integration error at all. This kind of gyro is big, heavy, and expensive, and they do drift over time. I think for new applications they are being replaced with ring laser gyros, which I believe are rate gyros. Mechanical gyros can have hard failures. I knew an F111 pilot that had experienced a gyro failure while in terrain following -- the airplane detected the failure and climbed very hard. He said it was very exciting.

I'm pretty good at making Kalman Filters, but as I posted in another thread, I have to refresh my memory of aircraft control.

[Unterhausen]

One sensor that hasn't been mentioned is a magnetometer which gives an orientation to the earth's magnetic field. They are noisy and slow, but they don't drift, so you can use them to correct for errors from gyros and accelerometers. At least that's the way I understand it without looking into it too hard.

[LukeZ]

I'm pretty good at making Kalman Filters, but as I posted in another thread, I have to refresh my memory of aircraft control.Say, does that mean you're volunteering to teach the rest of us? :D

I think there would be quite a bit of interest in the development of some Kalman routines, say in C, that we could use in our fixed wing autopilots. The AutoPilot project used Kalman filters but for helis, and apparently there is a difference. I've not waded through their stuff to a great degree yet, having put that portion of my autopilot off to last, and working on the easier stuff first. But it's looming and I'm wondering about it.

Assuming it's something I even figure out in the future, I would very much like to make the code open source in the same manner as the AutoPilot project - or perhaps it could be given to the Paparazzi folks. They keep talking about Kalman but have never gotten to it.

Well, any info you want to share with us I'll eagerly devour! :)

Luke

[vpatron]

Yes, I thought about it too. I have one of those handheld digital compass and it seems to work fine.

I found a sensor model "FGM-1". It's tubular 35mm long x 8mm diameter. I think you'd have to use 3 in orthogonal orientaions to get 3 axis reading.

Seems like the size and mounting constraints as well as stray magnetics from motors and high current wires are going to make this a no go for a tiny UAV?

-Vince

One sensor that hasn't been mentioned is a magnetometer which gives an orientation to the earth's magnetic field. They are noisy and slow, but they don't drift, so you can use them to correct for errors from gyros and accelerometers. At least that's the way I understand it without looking into it too hard.

[NewBee123]

I heard Kalman filtering uses estimation to predict future values using the present and past data. I may be wrong, but thats about all I know about Kalman filters. I would be really interested to learn them since I didnt do a good job at school when I took Filter design course. Here is a link with tons of info on Kalman:

http://www.cs.unc.edu/~welch/kalman/

If some one knows more links, please post them here.

Keep the info coming guys.... I am learning a lot from here.

[william541]

One sensor that hasn't been mentioned is a magnetometer which gives an orientation to the earth's magnetic field. They are noisy and slow, but they don't drift, so you can use them to correct for errors from gyros and accelerometers. At least that's the way I understand it without looking into it too hard.

Unterhausen,

I was thinking of the same thing and even ran some tests using a hall effect compass where I could read the raw A-D info in the x,y and z planes. I came up with a few formulas for determining pitch and roll using the earths magnetic inclination as an orientation reference, but got a bit lost in trying to solve all of the equasions in 3-d space. See the following starting at post #86 in this thread for what I was thinking about.

http://www.rcgroups.com/forums/showthread.php?t=267651&page=6&pp=15

If there are any trig wizzes out there, maybe it would still be a viable idea for short range aircraft in any weaher condition.

-Bill-

[Unterhausen]

Say, does that mean you're volunteering to teach the rest of us? :D
Luke

I'm looking at papers on Kalman filtering for fixed wing aircraft right now. I found 2 that looked interesting enough to print. Maybe I'll put together a thread about it.

[ALtitudeap]

unterhausen,
Please do. i have also found some papers but calculus was over 10 years ago for me so a thread and refresher could be good.

[sesat]

Unter, would you please post links to the two papers?

Ram.

[ALtitudeap]

Sesat,
Try here. http://www.cs.unc.edu/~welch/kalman/

[Unterhausen]

These are the papers I found just doing a google search for uav kalman filter
If you are interested in Kalman Filters themselves, there is an old paper called something like "How I learned to stop worrying and love Kalman Filters" that I liked. There is also a copy of the first chapter of Peter Maybeck's book online that does a good job of showing why you would use a Kalman Filter. If you really want to understand Kalman Filters, the problem isn't really calculus, it's the probability. But once you have the form, it's plug and chug.

http://www.google.com/search?q=uav+kalman+filter&hl=en&lr=&c2coff=1&safe=off&start=10&sa=N
This paper doesn't have UAV applications, but still looks good:
http://www.movesinstitute.org/Publications/extendedkalmanfilter.pdf

http://gge.unb.ca/Research/GeodesyGroup/tutorial/precision_navigation.htm

Here's a Master's thesis, haven't read it but it looks pretty good.
http://epubl.luth.se/1402-1617/2000/081/LTU-EX-00081-SE.pdf

http://www.acfr.usyd.edu.au/publications/downloads/2003/Kim197/FSR03-NavGuidCtrl-Kim

A paper about helecopters
http://www-users.cs.umn.edu/~stergios/papers/iros99_heli.pdf

This paper is a little light on details
http://www.u-dynamics.com/sensor_fusion/idc2002revised.pdf

Here's a paper from Crossbow that could use a few more equations.
http://www.xbow.com/Support/Support_pdf_files/Fusion_Filter_Algorithm.pdf

[lvspark]

Might find some source code here...???

http://www.google.com/search?q=%22Kalman+Filter%22+code&hl=en&lr=&start=10&sa=N

[ALtitudeap]

instead of Kalman Filter, What about PID Control.
http://www.parallax.com/html_pages/resources/custapps/app_bway.asp

[Unterhausen]

The code for a Kalman filter that will run on a PC is no big deal. A Kalman filter that would run on a AVR is a very big deal. Fortunately, aircraft dynamics are fairly slow.

Altitudeap,
There is a lot to be said for keeping things very simple. If properly calibrated, the sensors we can afford are probably good for short flights. If clever methods of negating the bias errors of the sensors can be devised, then a Kalman Filter may not be necessary. However, standard industry practice includes Kalman Filters, so I'm assuming they are necessary.
Kalman Filters are used for what is called an estimator, i.e. they tell you what the airplane and sensors are up to, and the actual control is done using PID. So they aren't necessarily competing techniques.

[vpatron]

If clever methods of negating the bias errors of the sensors can be devised, then a Kalman Filter may not be necessary. However, standard industry practice includes Kalman Filters, so I'm assuming they are necessary.
Kalman Filters are used for what is called an estimator, i.e. they tell you what the airplane and sensors are up to, and the actual control is done using PID. So they aren't necessarily competing techniques.

Unterhausen, so in layman's terms does this mean that GPS update is too slow (1 per second) so a Kalman filter predicts what the state of the plane is (roll, pitch, yaw, velocity) at some given time say 0.9 seconds after the last known state?

Regarding your first statement about a scheme for negating bias errors in sensors and not needing Kalman filters, can gyros be "reset" whenever GPS says the plane is flying flat? For example, can you assume 0 pitch when GPS says you've been at the same altitude for the past few seconds?

Just thinking out loud, I guess this can be error prone in situations like thermals where the pitch can be downward but altitude stays the same and also vice-versa in downdrafts.

Sorry for the simplistic questions, but it's been a while since I've been in school.

Many thanks,

-Vince

[Unterhausen]

Vince,
Those are really good thoughts. Using the GPS to update sensor error states is a really good idea. Actually, that's equivalent to using a Kalman Filter where you can trust the GPS, but not the sensors.

Generally, you control pitch, which is, as you say, different from altitude. One of the papers I looked at actually used multiple GPS recievers to estimate pitch and roll. So it potentially can be done. That paper is listed in the search link I gave above.