Comments on: TapeMeasure – Another Android experiment

By: tb

tb — Fri, 26 Aug 2011 06:09:00 +0000

I know this is an old post, but I found out a junior engineer at my company was forced (by an idiot boss) to attempt this. I actually have done a lot of work with MEMs accelerometers before they were ever put into phones, but was never asked. Please don’t get stuck like she did.
To put it bluntly, it’ll never work without correction from another sensor. Here’s why. Assume that ‘a’ is the true acceleration, but a real-world sensor measures error too. This error is included in the equation to remove gravity. This means that you partially leave gravity as real acceleration. This feeds back into the next step and the effect of that error increases as the square of time (due to double integration). So after say 10 seconds, it is multiplied by 100. After 20, by 400. And that is just the error from that ONE sample. All the others are accumulating on top of it with the same increase.

A second sensor like a camera can be used to zero out the error periodically and may be a good solution. Look up Kalman filters to see how this is done in inertial measurement units.

HTH, good luck.
t

By: Anilsharma Y

Anilsharma Y — Mon, 11 Jul 2011 12:15:00 +0000

All in all a bit of a waste of time )

By: CT

CT — Sat, 27 Nov 2010 22:24:39 +0000

I read buyer comments on Android that millimeters is not so good either.
Good exercise of exploration !

By: Stan

Stan — Thu, 19 Aug 2010 19:18:39 +0000

As oiok said Millimeters works fine: http://www.youtube.com/watch?v=iYpiVTdElPY

Stan

By: Leo

Leo — Thu, 24 Jun 2010 18:51:37 +0000

Hi,
This is interesting, have been looking at doing something similar myself…
A few comments:
I don’t know much about these sensors, but I don’t think there’s a need to subtract gravity as the accelerometer will only register if the phone accelerates (i.e. is dropped), in which case you want to know about it. This might be the reason you get a rapid distance increase. If the distance per second linearly increases, there’s your culprit. Also you might consider implementing a improved Euler or Verlet numerical integration, the second one stops computation error-caused silly increases in displacement in my experience. You’ll need one step of a simple numerical integration to get it started though.
Look forward to meeting you soon, Leo

By: oiok

oiok — Fri, 14 May 2010 21:05:12 +0000

Maybe you should take a look at Millimeters ( available on the Market: market://search?q=pname:com.droideilhan.millimeters ).