Hello everybody!Here's my setup:-Esky Honey Bee RC 4CH FP Ready-to-Fly Helicopter-Parallax Propeller-Memsic 2125 Dual-axis Accelerometer-GPS-A willingness to run the heli into the ground as many times as it takes :PAnd here's what I've learned about each of the above:HelicopterI realize this heli isn't very "strong" as far as thrust goes. I've been able to put a fairly big wireless camera on it but it only flew after removing the casing. This is my first helicopter so I'm going to have to make it work!Parallax PropellerThis is the new micro controller from Parallax, it runs at something like 12Mhz and it's got 8 cores, well documented, and well supported. I've run through the manual and can work my way through what ever code I need to come up with.Memsic 2125 Dual-axis AccelerometerI've gotten this thing running with a Parallax Basic Stamp(No prop as of yet) and it seems to work fairly well. It only detects 2 axis but I think thats all I'll need for the time(Roll and Pitch?)GPSThe only GPS I have is MUCH to heavy for the heli but I might consider buying a lighter one if need be.A willingness to run the heli into the ground as many times as it takeI mentioned that this is my first Helicopter. I haven't really even learned to completely fly the helicopter but I have crashed it many many times and each time I've picked up the pieces and gotten it flying again.And Finnally, my question: I've decided to start small, so I'm going to start w/ the accelerometer. How would I go about using the accelerometer to keep the heli balanced?I realize that say, if the accelerometer says the Heli is too rolled left I would just adjust the roll servo to roll the servo right but I don't know how I would determine how MUCH to adjust the servo... I could just do some trial and error and I'm afraid that there would be just so much error I wouldn't get any viable results. Is there's an equation I could use? Back to the trail and error idea, though. I thought that maybe I could design some kind of ball joint that would be under the heli and extend on a pole to the ground. That way the Heli doesn't have any chance to crash, it just leans on the ball joint. Thinking about it now.. the ball joint might actually need to be toward the center of roll/pitch so that the thrust doesn't simply correct for all error.Any ideas?
You need to be a member of diydrones to add comments!
I have just started on a project like this. I bought a trex 600 and I plan to use either an ap2000i or fs8 for flight stabilisation which should give me the stability for flight and for hover and for navigation a gps system which I have been using/testing on my aircraft. I know there is going to be a lot of problems to over come. I have a couple of questions though
Just to clarify If sitting in the hover with no input from the transmitter using the fs8 or similar when it rolls left,right front,back then stability program should correct this then? if you have the gps system switched on and one waypoint fixed in will the aircraft sit in a hover over this point?
I know there is going to be various forces acting on the aircraft and the only thing I can see so far happening is with the stabilisation program when it compensates for going in one direction it might overcompensate and then start a motion that it won't be able to correct.
I have found this off your main page which although there isn't much to it I might give it a go.
After the intitial challenge of getting the autonomous flight with my aeroplane I am now looking forward to the challenge of getting it to work on the helicopter and also trying to get the aeroplane which now flies itself to land itself
The goal is to measure acceleration in the horizontal plane parallel to the ground. The problem with dual axis accelerometers is that when your craft is tilted you are not longer measuring forces in the horizontal plane (no longer parallel to the ground). It is possible to resolve these forces if you know the angle at which you are tilted, hence the need for the gyros.
I have not worked with kalman filters but a member here has developed working code. This is a tough project but the information you need is out there.
Recommend dissecting 3 cheap rate gyros instead of the proto boards from Sparkfun. Unfortunately the Kalman filters don't get U much unless U have differential GPS to eliminate dynamic effects. 2 axis IMU's have been tried and don't do the job. U need a full quaternion attitude estimate to make sure pitch doesn't integrate into roll or yaw.
good description of the problem. likewise, I will be happy to see my craft maintaining attitude, even if it shifts around somewhat. heading and altitude locks are next, and then I'll worry about absolute position.
I'm sure someone here can point you toward a good kalman filter tutorial. another technique to consider is neural nets, which have a good track record in this type of application. that's actually the direction in which I want to head for data fusion, though it's pretty far down the list of priorities at the moment.
To make your craft hover you cannot assume no motion in the lateral plane, these are exactly the forces you have to measure and compensate for. In practice you cannot decouple tilt from acceleration in a helicopter. Additionally it is not possible for a dual axis accelerometer to distinguish between accelerations due to lateral movement or due to tilt. At the very minimum you will need a dual axis accelerometer and dual gyros. If you expect to maintain a constant altitude you will need a 3 axis accelerometer, and you will need a third gyro to maintain heading (or a compass). As mentioned previously the smallest of tilt angles will quickly lead to acceleration in that direction. This acceleration may be small in comparison to gravity, but over time your velocity will increase, and if uncompensated for altitude will decrease. Successfully measuring acceleration will allow you to calculate velocity and distance traveled. Your fancy code will have to do its best to keep all of them zero and you will be hovering at a fixed point in space. Good Luck.
Let me the first to say the obvious: you can't stabilize a heli with just an accelerometer. They can't tell the difference between gravity and inertial forces in a turn (both are accelerations). You need to bring in gyros, too, and then, well, you're really going to need that Kalman filter Propeller module code in the Parallax libraries. Basically, you've bitten off one of the hardest challenges there is in autonomous flight (there's a reason even full sized helis didn't have autopilots until just a few years ago). Just ask Jack Crossfire, and ace engineer (he posts here and at RCGroups) who's been working on this for a year, without success--it's hard!
You might want to turn stabilization over to a commercial unit, such as the FMA Co-Pilot, and design a navigation-only autopilot.
Replies
I have just started on a project like this. I bought a trex 600 and I plan to use either an ap2000i or fs8 for flight stabilisation which should give me the stability for flight and for hover and for navigation a gps system which I have been using/testing on my aircraft. I know there is going to be a lot of problems to over come. I have a couple of questions though
Just to clarify If sitting in the hover with no input from the transmitter using the fs8 or similar when it rolls left,right front,back then stability program should correct this then? if you have the gps system switched on and one waypoint fixed in will the aircraft sit in a hover over this point?
I know there is going to be various forces acting on the aircraft and the only thing I can see so far happening is with the stabilisation program when it compensates for going in one direction it might overcompensate and then start a motion that it won't be able to correct.
I have found this off your main page which although there isn't much to it I might give it a go.
http://vicacopter.com/
After the intitial challenge of getting the autonomous flight with my aeroplane I am now looking forward to the challenge of getting it to work on the helicopter and also trying to get the aeroplane which now flies itself to land itself
oh and yeah I cannot fly a helicopter at all
I have not worked with kalman filters but a member here has developed working code. This is a tough project but the information you need is out there.
I'm sure someone here can point you toward a good kalman filter tutorial. another technique to consider is neural nets, which have a good track record in this type of application. that's actually the direction in which I want to head for data fusion, though it's pretty far down the list of priorities at the moment.
You might want to turn stabilization over to a commercial unit, such as the FMA Co-Pilot, and design a navigation-only autopilot.