MEMS location tracking/recording integration with GPS


5 posts
by bantammenace2012 » Wed Oct 31, 2012 1:33 pm
GPS/Glonass needs to be able see a portion of sky to get a position but that is not always possible. Therefore my location tracking/recording requirement cannot be met by GPS alone.
I am considering developing a hand-held, field portable, sensor that can be used to precisely (+/- 20mm) determine where it has been wrt. the last GPS fix or to a fixed and accessible reference point. The output does not need to be real-time.

The vector would typically less than 10 metres in plan distance, 5 metres in height distance and would typically be less than 20 seconds in time from the last known GPS fix or re-initialisation on the reference point.

I was thinking of using something along the lines of the sensor fusion found in the Invensense 9 dof sensor. Ideally a time-stamped location vector with the maths (kalman filtering etc. ) done on-board would suit my purposes best.
What I don't know is whether:
a). the maths stacks up especially wrt. sensor drift ?
b). anyone knows of a software emulator that would help me tell if my desired precision is achievable within parameters defined above ?
b). anyone else is interested ?

I have read with great interest the DIY Drones stuff and I remember seeing a video of lots of micro drones flying in tight formation (it may have been a TED talk ?). They appear to be doing what I want but in reverse.

I did post in the Camera section to see if anyone knew of any possible MEMS functionality within the existing Broadcom SoC of the RPi (Some current smartphones do have such sensors) but I got no replies.
Posts: 55
Joined: Mon May 28, 2012 12:18 pm
by bantammenace2012 » Wed Oct 31, 2012 1:35 pm
My bad,
for MEMS please read MEMS IMU
Posts: 55
Joined: Mon May 28, 2012 12:18 pm
by Ravenous » Wed Oct 31, 2012 2:05 pm
Do you mean use the following device, in conjunction with GPS?

http://www.invensense.com/mems/gyro/doc ... -9150A.pdf

I suppose drift or noise would ultimately limit the precision you can get, so the only real way to see if 20mm accuracy is attainable is to build one and try it... does any device on the market offer such accuracy?

By the way I don't think the camera stuff would be help with this. From the above datasheet it seems to use an i2c interface, which various people are using on the raspi now.
Posts: 1228
Joined: Fri Feb 24, 2012 1:01 pm
Location: UK
by bantammenace2012 » Wed Oct 31, 2012 5:46 pm
Yes, thanks for the link.
I am not sure about cost or availability for very small numbers of such a chip.
Do Invensense offer evaluation kits like U-blox ?
Posts: 55
Joined: Mon May 28, 2012 12:18 pm
by hitcher404 » Sun Nov 04, 2012 9:00 pm
Not something easy to do with gyroscopes and accelerometers. To go from acceleration to displacement you have to integrate twice, which means errors/noise get amplified, even a small zero offset causes large erroneous measured distance.

Sebastian Madgwick (guy who wrote THE sensor fusion algorithm) got displacement from 6DOF imu for pedestrians accurately by being able to zero the reading each step:

http://www.youtube.com/watch?v=6ijArKE8vKU

Without some kind of point of reference like this every few seconds I don’t think estimating position to the accuracy you want is possible.
Posts: 11
Joined: Sat Jul 14, 2012 11:36 pm