maemo.org - Talk - View Single Post - Using accelerometer to know your position

	ndi	2010-05-10 , 17:14
	Posts: 2,050 \| Thanked: 1,425 times \| Joined on Dec 2009 @ Bucharest	#14

Originally Posted by juise-

To reliably detect all possible rotation with only linear acceleration sensors would actually require three sensors that are not placed on a same line.

With only two, you could position the device so that the axis passing through both sensors is aligned with gravity, and then rotate around that axis to make the device lose it's orientation.

If by "accelerometers" you mean the device that has 3-axis, like the one on N900, then you need two, as long as they are not aligned, as I said to navigate (I'll expand on that).

Your suggestion doesn't work because it's basically impossible to align and rotate a phone like that in real life. Also, in real life, rotation will show up on accelerometers because of centrifugal force.

You're right, in theory, it is possible to do that. In practice, I very much doubt it. For one, if the angle isn't 45 degrees (and it isn't), rotation will induce centrifugal on axis of each accelerometer at different rates because they have different angles to the axis, so the resultant will show different accelerations as left and right on each.

It will show up as rotation on the wrong axis. So, IMO, not worth it for a phone. Definitely a must for inertial navigation. Also, planes use gyros, not inducted weights. Also, they cost 900 N900s.

So, to summarize, you are correct, you need 3 to accurately detect all movement. But by the time that influences your navigation (it doesn't because navigation doesn't care you roll when navigating), you're hit by several other issues, related to precision, drift, sampling rate, etc.

__________________
N900 dead and Nokia no longer replaces them. Thanks for all the fish.

Keep the forums clean: use "Thanks" button instead of the thank you post.

Quote & Reply |

The Following 2 Users Say Thank You to ndi For This Useful Post:
juise-, theonelaw