The physics of making an iPad into an accurate aviation instrument | Hacker News Reader

The physics of making an iPad into an accurate aviation instrument

x-plane.com

46 pointsjoshwa26 comments

Yes, Austin Meyer (creator of X-Plane) is a bit of a loon, but if you can see past the ALL CAPS he does give some pretty good explanations of interesting aviation-related stuff.

More Austin Meyer goodness: http://www.x-plane.com/x-world/austins_adventures/

Building his own airplane (in incredible detail): http://www.x-plane.com/hardware/evo/evo.html

Incredible in-flight engine failure autoland device: http://www.x-plane.com/hardware/evo/9_seeker/Seeker.html

A rather convincing explanation of why very small jets just don't make sense: http://www.x-plane.com/hardware/evo/0_nojet.html

OK. I'm no expert on this myself, but for someone who claims to have just figured out how an attitude indicator works, the author certainly doesn't explain it very well. I was left with the impression that the author is uncertain of the distinction between an accelerometer and a gyroscope. An accelerometer, like he says, can be likened to a hanging weight or pendulum. Accelerometers can measure the direction of gravity of a stationary object because the only force acting on the pendulum is acceleration due to gravity. However, as soon as the object is moved, other acceleration vectors are added to the already-present gravity vector.

(Imagine an accelerometer hanging from a rope. If you swing that rope in a circle around your head, the accelerometer will report that the force vector points to its bottom, away from your head, and not down to your feet as gravity alone would indicate.)

Which gets me to the OP. Because the article does not make a single mention of angular momentum (among other reasons), I would guess that the author doesn't understand how gyroscopes work. A gyroscope is essentially a spinning disc-shaped mass that resists change to its axis of rotation. If you built a housing that allowed it freedom of motion in three dimensions (called a gimbal (http://en.wikipedia.org/wiki/Gimbal)), it would remain on its original axis no matter how you tilted its enclosure (barring the inevitable energy loss due to friction, etc.)

So, in the case of an airplane, you would set the disc spinning on level ground. Then, while you're flying, any changes to your pitch or roll would be around the gyroscope (which, remember, is allowed to move freely relative to the cabin). The gyroscope wants to preserve its angular momentum, so it will still be spinning on the same axis relative to level ground. Any difference between the gyro and the cabin therefore indicates the orientation of the plane relative to the ground.

You'd probably get a more concise explanation from the Wikipedia page on Gyroscopes (http://en.wikipedia.org/wiki/Gyroscope). The Wikipedia page on attitude indicators (the visual display for a cockpit gyro) is also quite coherent and less prone to rambling (http://en.wikipedia.org/wiki/Attitude_indicator).

[EDIT: Toned down some grumpiness. I must be tired.]

Private pilot here. Mechanical attitude indicators don't need to be reset on level ground, when they first start spinning they're almost always tilted but once they spin up they level out on their own. Because they're driven by vacuum from the engine, sometimes they don't fully spin up and level out until the engine's being run up or takeoff.

I've always wondered how the attitude indicator self levels, and the OP's explanation sounds plausible. Next time I go flying I'll have to spend a few minutes in a constant rate turn to see if I can make the attitude indicator drift like he says it should.

Edit to add, here's a link to a youtube video showing the inner working of an AI, including the pendulous vanes, neat:

http://www.youtube.com/watch?v=KUSklh3MKtA

Yeah, I kind of glossed over the vacuum aspect of the attitude indicator because I don't understand all its intricacies. The video pretty much confirms my understanding of how the gyro is kept spinning. The original author's description of air puffs sounds a lot like how the filtered air is ducted in to the housing and vented off in quadrants. I suspect that has more to do with maintaining the gyroscopes precession (http://en.wikipedia.org/wiki/Precession) than leveling the gyro in mid-flight.

I've heard that during prolonged banks or pitches, the AI can eventually find a different "level" than the true horizon. I'm trying to find more info into that phenomenon though.

In the mean time, I've found this article that seems to be talking about the same accuracy and drift issues as the original post, but explains the issues with more clarity and also includes sensor graphs and code samples: http://myahrs.wordpress.com/2012/04/24/turning-errors-contin...

The system you're describing (which is a real thing) would allow you to measure your absolute orientation.

The gyroscopes in iDevices only measure rotational acceleration, which is only the second derivative of your orientation, so you need to do a bit more work to extract an (slightly error-prone) orientation out of it.

I haven't specifically used an iDevice, but all the gyroscopes I'm familiar with provide the first derivative, not second.

I'm also curious why he doesn't mention any use of magnetometers, which can give absolute orientation (not just the derivatives). Are the changes in position/elevation too large for airplanes to rely upon the local magnetic field of the earth?

Any difference between the gyro and the cabin therefore indicates the orientation of the plane relative to the ground.

Unless you flew far enough that local level ground is no longer approximately parallel with level ground where you took off.

There is a difference between electronic gyros and mechanical gyros.

Could you enlighten us? I've read about fiber optic gyros, but I assume they haven't gotten small enough to fit in a mobile device. (Since they require kilometers of fiber.) I really don't know enough about how the solid state (?) gyros work.

Though this is the internet....

Far too many hyperbole. Much of this is straightforward when dealing with GPS data. And all the shortcomings of relying on gyro data can be found by searching for dead reckoning systems.

Don't get me wrong, I am very impressed with the x-plane application, but the writing style assumes a complete idiot.

I agree. Some paragraphs in I was thinking that the solution is to just calibrate it when it's on the ground, but was curious what solution he came up with...

> While on the ground, use the ACCELEROMETERS to see which way is up.

Oh. Yeah, I guess I should have seen that coming.

Still, the more people write this enthusiastically about a topic considered as "mundane" as this by an average "complete idiot", then I'm all for it.

Because despite the writing style, I could feel the excitement in this, and it was worth it for that alone.

And it has made me more excited about this topic.

He continues on, saying that [ground calibration] doesn't work for very long due to errors in the reported rates of change.

Put another way, holy crap that was nearly impossible to read.

Man, that's exactly how I felt. I think we need a tl;dr for this -- it was incredibly frustrating.

I couldn't read due to the much caps locks. After reading the comments, I thought I might add a few things.

As others have pointed out, outdoors you would use mostly GPS, and maybe a camera pointing to the ground to estimate the position. Accelerometer are more or less worthless when you want to estimate a translational position of a free flying rigid body. However, orientation can bes estimated very well with the sensors you find in a smartphone. At least outdoors.

The acceleration gives you the orientation of the earth gravity vector, assuming that translational acceleration can be neglected. This information corresponds to pitch and roll in an airplane. The iPad, and most smartphones, also have a magnetometer (digital compass) which gives you the direction of the earth magnetic field and which works reasonably well outdoors. You can combine these two 3D vectors with an iterative algorithm, e.g. QUEST [3] or Madgwicks AHRS [4] to obtain an estimation of your orientation in quaternions. No gyroscope required so far. Usually, the gyroscope is only used to give a more accurate angular velocity and acceleration signal, speed up the orientation estimation and serve as a short term replacement in case of a larger disturbance in acceleration or the magnetic field. A Kalman filter is the most common way to do this.

If you want to play around with these values on your phone I recommend the Sensor Kinetics [5] app for Android, or if you have an iOS device, I believe XSens has an app, too. There is usually need to implement the above method yourself. Most IMU's you can buy of the shelf provide these higher level estimations already.

[1] http://en.wikipedia.org/wiki/MEMS_gyroscope#MEMS_gyroscope [2] http://www.apple.com/ipad/specs/ [3] e.g. EXTENDED QUEST ATTITUDE DETERMINATION FILTERING Mark L. Psiaki 1999 http://www.google.ch/url?sa=t&rct=j&q=&esrc=s... [4] Includes working (tested it myself) Matlab, C and C# code http://www.x-io.co.uk/node/8 [5] https://play.google.com/store/apps/details?id=com.innoventio...

"If anyone thinks that they are revealing new information to me by telling me the difference between centrifugal force and acceleration, I will fed-ex a crazed baboon that is high on crack to their house."

I can think of a few people I'd like to see receive a package from fedex containing a crazed baboon high on crack.

He is being sued by Uniloc, I think this is the same suit against Mojang (Minecraft)

http://www.x-plane.com/x-world/lawsuit/

previously: http://news.ycombinator.com/item?id=4523380

XCSoar is a very nice free and open source flight computer that includes an artificial horizon. It runs on your Android phone, old PDA, and many other devices.

http://xcsoar.org/ https://play.google.com/store/apps/details?id=org.xcsoar

It's designed for sailplane, hang glider and paraglider pilots, and is slowly disrupting the incumbents in this space. Source code is C++.

It seems that the author developed the underlying kinematics needed to implement an inertial measurement unit (src: http://en.wikipedia.org/wiki/Inertial_measurement_unit ), though with his introduction of GPS, he could probably convert it to an inertial navigation system (http://en.wikipedia.org/wiki/Inertial_navigation_system). I wonder if the author is aware of the Extended Kalman Filter (src: http://en.wikipedia.org/wiki/Extended_Kalman_filter ) -- he could fuse the noisy sensor data with the aircraft's kinematic model to cancel out some of the error he was mentioning getting. The best resource I've found online about how to use Kalman filters and EKFs is course notes from a simultaneous localization and mapping workshop taught at Drexel University in 2006 (src: http://prism2.mem.drexel.edu/~billgreen/slam/slam.html ).

There are external GPS receivers specifically to improve iPad GPS performance in aviation.

For example: http://www.mygoflight.com/external-gps/

The writer would benefit greatly from the Ardiuno community. Most of this stuff is open-source code. Quite possibly, some of the authors solutions were learned from it. However, no credit was issued for where the author learned most of their lessons.

I do appreciate the layperson's explanation of an attitude system applied to a smartphone/tablet. Many do not realize just how powerful these mobile devices actually are.

The Arduino community did not invent the Kalman filter or sensor fusion.

No, but they have aggressively applied them to self balancing robots which have many of the same issues as the author is dealing with (isolating actual up vector from acceleration vectors etc)

And they're not the only ones to have done so. Adruinos aren't the be-all and end-all of embedded systems, and most projects do not need to give them any credit whatsoever.

I couldn't make it to the end of the article; the writing style is too frustrating.

Reply on news.ycombinator.com