Printable View
I decided to start documenting the project I am working on - an autonomous quadrocopter (helicopter with 4 propellers) in the form of a log/tutorial.
Feel free to stop by if you are into these kind of things!
http://cyberfish.wecheer.com/quadrocopter
I haven't really gotten to the programming part, though. Most of it is still electrical/mechanical stuff. But there will be a lot of programming later on.
Looks like a fun project.
I was going to use a Cortex-M3 (STM32), but later decided to go with a gumstix (single board computer with a Cortex-A8) because I needed more memory (among other things).
Yah edited my post, apparently its too late for me to read atm....
Thought about using a mobile phone to send text-messages with commands to the copter? IE go to gps-coordinate, stay at altitude, take pictures every x seconds and mms them back to the phone and stuff like that? Would be cool so you dont have to rely on the radio or wifi. I realize this is very much in the future but could be a nice addition if you succeed with the radio and wifi.
It's an interesting problem: orientation in space. I wouldn't worry about accelerometer chips not being perfectly perpendicular. You can apply a matrix transform correction in software.
How about some sort of compass and a sun angle detector? I realize a solid state fluxgate compass may be problematic in a noisy environment like a copter platform. But I've seen sun angle detector modules... just an idea that may be neater than gyros.
Perhaps some level digitizers on 2 axes, plus the sun angle data. As long as the platform is guaranteed horizontal, the sun's angle should be all that is required to get orientation. Now for altitude, there may be accurate barometric gauges.
Be advised that the type of parts you're getting are getting into scary territory that may set off alerts from homeland security. That's just my opinion. Maybe I'm overreacting.
Anyway, sounds like a nice project. I wish I was there to help. If I had any time for hobbies any more.
I actually thought about that. Using cellular network for internet connection. Speed should be adequate. The problem is, I couldn't find a cellular chip that will connect to 3G, and is available to hobbyists. GPRS/2G speed is too slow for video streaming.Quote:
Thought about using a mobile phone to send text-messages with commands to the copter? IE go to gps-coordinate, stay at altitude, take pictures every x seconds and mms them back to the phone and stuff like that? Would be cool so you dont have to rely on the radio or wifi. I realize this is very much in the future but could be a nice addition if you succeed with the radio and wifi.
Otherwise it would be a really cool idea!
Yeah the perpendicular thing is not really a big problem. We can probably just live with the tiny error, too.Quote:
It's an interesting problem: orientation in space. I wouldn't worry about accelerometer chips not being perfectly perpendicular. You can apply a matrix transform correction in software.
How about some sort of compass and a sun angle detector? I realize a solid state fluxgate compass may be problematic in a noisy environment like a copter platform. But I've seen sun angle detector modules... just an idea that may be neater than gyros.
A 3D compass can possibly work, assuming they exist, but I'm not sure if it will give me enough precision and update rate (at least 20Hz minimum to make it stable).
Very cool.
I'm especially fascinated by the Gumstix (after looking it up :P).
I'll be following your progress. :)
It's amazing how much they can cram into a computer that small. It looked even smaller in reality. About the size of an AA battery.
The speed is like that of a PC from 7-8 years ago (Pentium 3 era).
Only draws about 2W, too (with wifi off).
Update: mechanical design done.
You should check out rcgroups if you haven't already. They had some very interesting designs, plus videos of how theirs got destroyed.
Maybe there's more to your plan than I know about, but your frame design looks terrible.
-> You have a lot of mass there that doesn't need to be. You should streamline you design such that there isn't a bunch of mass around the outside, it doesn't help. Oh and just make the bottom layer a little thicker and forget the top one.
-> It's flat. You should locate the blades at the top, and keep the 'platform' under the blades, not in the middle, this will help a lot with stability.
I was hoping to go for high strength (crash resistance), hence the relatively thick (~2cm) and complete outer rim. I don't think it will add considerable mass.Quote:
-> You have a lot of mass there that doesn't need to be. You should streamline you design such that there isn't a bunch of mass around the outside, it doesn't help. Oh and just make the bottom layer a little thicker and forget the top one.
My rationale for having the blades in the middle is to give them some protection (if it ascends or descends into something), but you are right, I haven't really thought about aerodynamics. I will remove the top platform if it significantly impacts stability.Quote:
-> It's flat. You should locate the blades at the top, and keep the 'platform' under the blades, not in the middle, this will help a lot with stability.
Where is the gun turret?
Ah, I didn't think about think about the durability. Makes sense now.
It's hidden. You need to type the cheat code "WEAPONS HOT".Quote:
Where is the gun turret?
same with god mode, right?
how do you control the yaw ?
all i can imagine it doing is going spinning out of control. you have four engines all turning in the same direction you are bound to have some serious torque issues. in the same manner the tail feathers spinning on a helicopter to counter act the main rotors torque. spinning and spinning and spinning then pounce on them "ouch!" those toy helicopter rotors sure can hurt <do not ask>.
i get how the quadhelicopter pitches forward and back and rolls left and right but not yaws left and right.
suggestions. static dihedral or static engine offset. have few more but those should do.
why four engines ? less weight with three and less power consumption. weight against the thrust lift of four ?
should it go spinning out of control what is the programming function that you will call ?
meowie ?Code:bendover( &kiss , " tail feathers good bye!" );
2 rotors spin one way, 2 spin the other.
Quadrotor - Wikipedia, the free encyclopedia
There is independent control of pitch, roll, yaw.
3 propellers requires variable pitch hence mechanical linkage (if I remember correctly), and actually consumes MORE power. Props have to spin faster to generate the same lift, and brushless motors are more efficient when they are slow. Quad is also more stable.
And are you going to put this on youtube when you are done so we can see it.
For sure if I ever get there. It's a very challenging project right at the limit of my ability... so I'm not all THAT confident.
It will probably take at least another month or 2, though.
You should be using 8 props, that would be cool. (and probably more stable) :D
Yeap, you can just keep adding more and more props :).
The algorithm gets more difficult with more than 4, though, because with 4 you have independent control over all axises already. With more than 4 it becomes over-defined (or whatever the right term is).
True, but I don't think 8 would be much harder at all.
Still only read those 2 axis. With 4 'primary' fans on the axis, and the other 4 'secondary' fans on the corners. Because all the secondary fans would be equidistant from it's neighboring axis, simply make each primary fan's manipulation take exactly half the effect on it's neighboring secondary fans. Basically, each secondary fan would be operating at the average speed of it's neighboring primary fans. Oh, and to do it this way, you would need all the fans in a circle, not square-like, naturally.
Pretty simple, really. :)
Neighbouring props spin in opposite direction. Average would be close to 0 :).
If we just make it spin one direction it will ruin yaw balance (it will start spinning). It's not that simple. There are 3 axises of rotation to think about, and all props have effect on all 3 axises.
Okay lets say we're looking from the back side of the craft.
A gust of wind tilts it to the right. So your auto stabilizer system responds by speeding up the right prop to 1500 rpm, and your left prop down to 500 rpm, while the front and back maintain the hovering speed of 1000 rpm. The 4 props I've talked about all spin clockwise. Now, the front-right and back-right props would respond by spinning at the 1250 rpm, and the left-front and left-back props would respond to 750 rpm, these 4 would be spinning counter-clockwise.
This way the non-absolute average of all 8 props comes to 0 rpm, like you want.
That might work. Now you just need to convince me it's better than 4 :)
8 is always way cooler than 4 (look at spiders) - and cooler is better. :p
In (more) seriousness: By having more props, your system is more fault tolerant, so if one of the props, for example, malfunctions, then your less likely to wreck. You could also carry more weight, and because of it's mass increase, it'd be less sensitive to atmospheric disturbances. Heavier means more room for faster computer equipment, and longer range radio equipment. If you don't want to take advantage of that extra room for weight, then you got yourself, an even more stable craft.
Considering, I think the added coolness and stability would be worth it. :D
Fault tolerance is the biggest advantage I believe, but is achievable with just 6 (any 1 motor can fail without losing control of any axis). 5 also gives some tolerance (if 1 motor fails, you lose yaw control, but still have pitch and roll, so you can hold altitude and do a controlled landing). But of course, that would also require weird algorithms.
As for weight... I'm actually trying to go the other way (make it lighter). My current design would weigh ~700g already. That should be enough for stability. I don't think stability (in terms of against wind) is as important for choppers as it is for airplanes, because planes have large areas for wind to push on (by design), and choppers don't. Any more weight would just eat battery life, and require bigger motors/props. Less weight also allows faster acceleration, which actually means better stability (the motors can react to changes faster).
I don't need any faster computing eqeuipment. I have a 600MHz ARM onboard already. That's way faster than what I need (stream video). The control algorithm will be limited by the relatively low update rate of the IMU (sensor that gives orientation information).
that would have been one of the several other suggestions. i could not work out how it would work. i could only foresee it getting stuck in a correction control loop. ie pitch forward a few degrees and it would have to correct for yaw. correct for the yaw and it would roll. then it would have to correct for the roll. while doing so it would offset the pitch. and so on.... i almost went catatonic trying to work it out.
less weight would be less stable in gusty wind.
Nope. INDEPENDENT control means you can change any one without affecting others.
I suggest Google if you want to learn more about it.
if you were sitting in the quadhellicopter for example engine one would be on the left side. engine two would be infront. engine three would be on your right side. engine four would be in back. engine one rotates clockwise while engine three rotates counterclockwise. if you increase the rpms in engine one you get more torque therefore you get yaw. if you increase the rpms in engine one you get more thrust therefore you get more lift therefore you get roll. by increasing the rpms in engine one you get both torque and thrust therefore you get both yaw and roll. independent control may mean you get only one axis but i do not see it happening here meow as explained above. if engine three decreases in rpms the offset only becomes more sensitive and happens faster.
Quote:
Nope. INDEPENDENT control means you can change any one without affecting others.
get it flying yet ?
No, see the diagram for rotation direction.Quote:
for example engine one would be on the left side. engine two would be infront. engine three would be on your right side. engine four would be in back. engine one rotates clockwise while engine three rotates counterclockwise
Quadrotor - Wikipedia, the free encyclopedia
If you increase front engine and decrease back engine by the same amount, you get pitch and nothing else.
If you increase left engine and decrease right engine by the same amount, you get roll and nothing else.
If you increase both front and back and decrease left and right all by the same amount, you get yaw and nothing else.
If you increase all engine by the same amount you get lift and nothing else.
took a closer look at the engines configuration on wiki page. so it is like kirkoff where everything has to equal 0. that is what was missing one increases in rpm while other decreases by the exact same amount with that config.
you need alt so you do not hit ceilings or go real high and run out of battery and come crashing down. bar would work. so you also know where gnd is. set 0 before lift off. landing pads needed ?
there is one other thing that you may want to consider adding to your quadhelicopter robot. that is a dmu <distance measuring unit>. why ? if you set the alt to 0 and it take off one foot up then goes over a cliff you alt will read 1 while there may be hundreds of feet below or no close ground below or near. totally optional. most mobile robots have some type of switch sensor at the bottom so they do not go over stairs. like a cut off switch so when it hits or releases it flips the switch depending on how it is set up. for something flying a dmu would be better. there are several types. ultrasonic is the most common. meow. used mostly for landing.
Thanks for pointing out the Gumstix modules, cyberfish. I'm enthralled by the Overo Fire and I'm particularly interested in the package including the Chestnut 43. It looks like you could almost build your own touch screen mobile phone with that kit, except of course it doesn't include a GSM radio.
I'm an ARM junkie, and I'm excited there are things like this out there.
Yeap, thought about it. There are pros and cons between infrared and ultrasonic, but I'm pretty sure either will work. I'm not planning to do automatic landing at this stage, though, so that can wait.Quote:
there is one other thing that you may want to consider adding to your quadhelicopter robot. that is a dmu <distance measuring unit>. why ? if you set the alt to 0 and it take off one foot up then goes over a cliff you alt will read 1 while there may be hundreds of feet below or no close ground below or near. totally optional. most mobile robots have some type of switch sensor at the bottom so they do not go over stairs. like a cut off switch so when it hits or releases it flips the switch depending on how it is set up. for something flying a dmu would be better. there are several types. ultrasonic is the most common. meow. used mostly for landing.
That would be really cool.Quote:
Thanks for pointing out the Gumstix modules, cyberfish. I'm enthralled by the Overo Fire and I'm particularly interested in the package including the Chestnut 43. It looks like you could almost build your own touch screen mobile phone with that kit, except of course it doesn't include a GSM radio.
I'm an ARM junkie, and I'm excited there are things like this out there.
GSM modules are available, too (SparkFun Electronics - Cellular) if you don't mind some soldering. But I don't think any of the ones available to hobbyists support 3G, yet. You'll probably also need to write a driver for it, though I think some GSM chips/modules have very simple interface that just takes simple messages (the module does a lot of things for you).
You can use a magnetometer for orientation. My employer was using one in conjunction with an accelerometer to avoid having to use a gyroscope in order to decrease power consumption for a different, non-helicopter project (communication was with zigbee).
I've got a main/tail rotor helicopter that also uses a brushless motor as the prime mover for the main rotor. I'm always adjusting to the gain on the gyroscope and/or fiddling with the trim for yaw on the radio control. The gyro is sensitive to temperature and humidity, but I'm not sure how sensitive a typical magnetometer is to temp. and humidity.
Theoretically, with magnetometer (which gives you yaw), and accelerometers (which give you roll and pitch), you can determine your orientation.
The problem is, and it's specific to choppers, there is a lot of vibration, and that introduces a lot of noise in accelerometers.
That's why we integrate (mathematically) gyroscopes to get accurate orientation, and accelerometers to correct the drift (inherent in integrations) over time.
I've been watching this for a while now.
I have to say, I'm so very jealous. ;_;
Soma
What do you mean by "accelerometers give you roll and pitch."Quote:
Theoretically, with magnetometer (which gives you yaw), and accelerometers (which give you roll and pitch), you can determine your orientation.
As I understand, the magnetometer will give you an orientation directly (yaw), it's just less accurate than using a gyroscope (but uses less energy!)
True. I suppose if you've got the onboard computing power, you can filter out the noise inherent from the helicopter propulsion system. I suppose one may accomplish this by recording the signal produced by running the chopper tethered in some appropriate fashion, recording the signal (from the accelerometers) and then while in flight pay less importance to those frequencies of vibration when analysing the signal from the accelerometer.Quote:
The problem is, and it's specific to choppers, there is a lot of vibration, and that introduces a lot of noise in accelerometers.
Another approch is to use two differently sized accelerometers (as each would theoretically have a different resonating frequency).
[quote
That's why we integrate (mathematically) gyroscopes to get accurate orientation, and accelerometers to correct the drift (inherent in integrations) over time.
[/quote]
Yeah.
Ah, by orientation I mean yaw, pitch, and roll. I need all 3 axises of rotation because the computer needs to stablize the chopper. Pitch and roll are actually more important than yaw for me.
For yaw I am using GPS, but many people use magnetometer.
For pitch and roll, accelerometer with low pass filter may have too high latency. For stable flight, the computer needs accurate orientation (in 3D) data about 10 times a second.
Gyroscopes give very good short term accuracy.
I guess I still don't know how you intend to (directly) use the accelerometers to give orientation as accelerometers measure linear accelerations, although placing them at a known position relative to your craft's center of gravity may be an interesting approach.
In general, it seems like you're describing this type of device, an IMU (Inertial Measurement Unit), where this one has a vector accelerometer and a dual axis gyroscope (but, doesn't give yaw, as you mentioned):
SparkFun Electronics - IMU 5 Degrees of Freedom IDG500/ADXL335
Gyroscopes don't just give short-term accurate results. You are right, there's error inherent because the only way the gyroscope could perfectly keep its orientation is if it had infinite angular momentum (accuracy is proportional to power consumption). But, the error can be taken into account, and will be described in the data sheets.
I'm not sure about the low pass filter causing too high of a latency, and I believe the gyroscopes will have some sort of low pass filter built in, whether you like it or not (see the schematic on this data sheet):
http://www.sparkfun.com/datasheets/C..._Datasheet.pdf
I would also be very careful about assuming that you need to have a hundred readings a second from the gyroscope. For one, the propulsion system won't respond that fast anyway, and two, it becomes difficult to write an adequate control system, especially for a helicopter.
I know a bit about control systems for maritime applications. The oil rig that blew up in the Gulf of Mexico, the Deepwater Horizon, was what's called a DP MODU (Dynamic Positioning MObile Drilling Unit). The smaller vessels that bring supplies to such rigs also have a DP system, such that it can remain along side the large rig while transferring cargo.
If you hadn't heard of it already, the response amplitude operator and its associated links may be helpful for the control system part:
Response amplitude operator - Wikipedia, the free encyclopedia
I guess I still don't know how you intend to (directly) use the accelerometers to give orientation as accelerometers measure linear accelerations, although placing them at a known position relative to your craft's center of gravity may be an interesting approach.
In general, it seems like you're describing this type of device, an IMU (Inertial Measurement Unit), where this one has a vector accelerometer and a dual axis gyroscope (but, doesn't give yaw, as you mentioned):
SparkFun Electronics - IMU 5 Degrees of Freedom IDG500/ADXL335
Gyroscopes don't just give short-term accurate results. You are right, there's error inherent because the only way the gyroscope could perfectly keep its orientation is if it had infinite angular momentum (accuracy is proportional to power consumption). But, the error can be taken into account, and will be described in the data sheets.
I'm not sure about the low pass filter causing too high of a latency, and I believe the gyroscopes will have some sort of low pass filter built in, whether you like it or not (see the schematic on this data sheet):
http://www.sparkfun.com/datasheets/C..._Datasheet.pdf
I would also be very careful about assuming that you need to have a hundred readings a second from the gyroscope. For one, the propulsion system won't respond that fast anyway, and two, it becomes difficult to write an adequate control system, especially for a helicopter.
I know a bit about control systems for maritime applications. The oil rig that blew up in the Gulf of Mexico, the Deepwater Horizon, was what's called a DP MODU (Dynamic Positioning MObile Drilling Unit). The smaller vessels that bring supplies to such rigs also have a DP system, such that it can remain along side the large rig while transferring cargo.
If you hadn't heard of it already, the response amplitude operator and its associated links may be helpful for the control system part:
Response amplitude operator - Wikipedia, the free encyclopedia
EDIT:
The math looks a little bit intimidating, but, if you've taken a course in differential equations, the RAO theory assumes that everything in motion can be described by a second order system. The RAO basically tells you how the vessel is going to respond to a given driving function (in the case of your quadrocopter, the driving function will be, say, the forcing from wind currents). Note that the forces are described in the frequency, not time, domain (waves!)
Accelerometers can measure both dynamic acceleration and static acceleration (due to gravity). That's a very common approach.
I am using a 6 DOF IMU.
SparkFun Electronics - ArduIMU+ V2 (Flat)
Hmm. Gyroscopes output a voltage proportional to angular velocity (how fast it's spinning). They are available with different resolutions. If we integrate that, we have orientation (pitch and roll).Quote:
Gyroscopes don't just give short-term accurate results. You are right, there's error inherent because the only way the gyroscope could perfectly keep its orientation is if it had infinite angular momentum (accuracy is proportional to power consumption). But, the error can be taken into account, and will be described in the data sheets.
The error comes from integration. Since we don't have infinite sampling rate, the integration will drift over time. That's why we use accelerometers to correct it. In theory, they produce redundant data. In reality, one is good for short term accuracy, and one for long term accuracy. We combine the 2 inputs with something like a Kalman filter.
The control loop will most likely run at 10Hz, and I'm pretty sure it's not possible to design a filter (digital or electrical) that will get rid of the noise, and still have good enough response time. On the other hand, gyroscopes have almost no short term noise, as they are not affected by vibration. So no filtering is needed on the gyroscopes. They may include one to filter out things higher than the Nyquist frequency, but the bandwidth of that would be much higher. According to the datasheet of the one I'm using, it's 140Hz.
My brushless motor controllers can change speed every 20ms (50Hz), but going that high would be overkill.
I guess I thought you were talking about using accelerometers for orientation, but it's moot.
Do you have the rest of the equipment already purchased? As much fun as I'm having crashing my traditional tail rotor copter into my shins* I'd much rather build a quadrocopter.
I saw on TV a homemade helicopter with each 'pod' of the 'quadrocopter' having an upper and lower propeller, the purpose being that they could change the orientation by increasing/decreasing speed but without compromising lift. It seemed like a hack job to me, but when they actually flew it, well I've never seen such a light heli so stable.
*I'm not actually having fun crashing it into my shins.
I AM talking about using accelerometers for orientation. But we have different definition of orientation (I think). By orientation I mean 3D orientation (pitch, roll, yaw), and I'm guessing you mean 1D orientation (yaw).
I've bought most of the parts. Keeping a detailed log on my site, too.
http://cyberfish.wecheer.com/quadrocopter
I'm done the mechanical design. Working on the circuitry now.
The upper and lower propeller design is good for stability, but EXTREMELY power inefficient (due to one propeller blowing into another), and will run your battery dry very quickly.
And as you say on your site, accelerometers measure linear acceleration, not rates of rotation. I haven't read further into your design, but what I can tell you from contract work my employer exposed me to, you're not really going to be able to 'determine orientation' from acceleration. The project I am referring to was a weather buoy project, that used accelerometers to transmit information about sea states. The problem is that as a buoy pitches, the accelerations have to be corrected for orientation (not the other way around!). The purpose of using a magnetometer to correct the accelerations for orientation was to decrease power consumption and cost.
Regardless, the IMUs you've listed seem to package this all for you.
EDIT:
But, I do want to say thanks for keeping track of the parts you've bought. I had started doing the research to build something similar and basically got bogged down with the many options out there. All the parts you'd found seem very reasonable as far as specs and cost.
If you have an object that's not accelerating (stationary or not), the 3 axis accelerometers will output a vector of length 9.81m/s^2, pointing to ground.
From the direction of the vector, we can calculate our pitch and roll. The only part of orientation we can't figure out is yaw, so I use GPS for that. Some people use magnetometers.