How do I figure out the orientation of a accellerometer?

It's no at all easy to work out when a plane leaves the ground or lands just by acceleration. You could probably detect the take off by looking at acceleration along the ground but it would be very easy to get false triggers for landing.

You would need to record lots of flight data and then devise a way of processing the data to correctly identify what was happening. Adding a pressure sensor might help (if it's in the hold).

It is easy to work out the orientation of a DC coupled accelerator if it isn't being accelerated (it might be hard to know this)  - but you can vector sum the readings from the XYZ axes and this (when the accelerometer is at constant velocity) will give you the angle of the earth's 1g gravity relative to the accelerometer.

The accelerometers and gyros in  MPU-6050MPU-6050 are not good enough to do classical inertial guidance but they can work in sensor fusion type applications to give useful information InvenSense may have some applications data that will help otherwise expect to do a lot of Googling and reading

MK

Thank you Michael!

I had an idea that this wasn't gonna be an easy task, and I also had the idea that it would take a lot of research and flight data.

I travel often, so hopefully I could make a tiny "beta" device that would track and log the data on and SD card, but for that, I would need to know how to get the correct data in the first place!

So I will try and look into what you wrote, and see if I can figure something out to do this for me!

I think you could probably use accelerometer data and write an algorithm to recognize something like a plane take off.

If you are recording constantly and detect a period of consistent acceleration in one axis (the point where the plane accelerates on tarmac) and then sometime after that you detect acceleration in a different axis (the point where the plane leaves the ground), then you could pretty definitely label this as a takeoff event.

If the Arduino/MCU keeps polling during flight and detects a pretty significant increase in g force, it could like the force of the plane touching down again. You'd just need to do some thresholding to make sure that turbulence during flight doesn't throw off your sensor.

To answer your other question, determining position in space from accelerometer or gyro data is called dead reckoning. I've done this a bit with the Genuino 101, which has built in functions to basically do this for you. Doing this from scratch is a fairly involved process, because you need to figure out some algorithms for repeatedly doubly integrating the acceleration data to obtain position data and quaternion data. The Genuino 101 has built in accelerometer and gyroscope so that can basically prebuild the orientation functions. I tested this a bit when I reviewed the Genuino 101 for a roadtest . Depending on the Arduino you use you may need to find some Arduino example code to get started.

Hope this helps!

Jordan

Interesting problem.

The only constant acceleration is gravity, all other accelerations are transient, so when the acceleration doesn't change for long periods of time it means the bag is stationary or at constant velocity.

After takeoff the plane climbs for many minutes at an angle of 10-15 degrees. This is much longer than any conveyor belt trip, so it is fairly easy to detect. But it means the steady state (cruising) will take at least 45 minutes to stabilize. Sitting on the tarmac before takeoff will also be much longer than any baggage handling accelerations, so you will know which way id down before takeoff.

On landing aircraft decelerate at 3-6 knots each second (about 1/4 g) for 10 - 20 seconds, which is longer than any turbulence, so it should be detectable. It is also in the opposite direction from the takeoff acceleration.

An alternative is just to set a timer for the ETA to turn the radio on.