Neo1973 Update!

Ian Stirling OpenMoko at mauve.plus.com
Sun Jun 17 16:22:51 CEST 2007


Lars Hallberg wrote:
> openmokolist.50.minime at xoxy.net skrev:
> 
>>> the GPS daemon would need to be updated to allow for kallman filtering
>>> using those accelerometers, so that the GPS apps could continue pointing
>>> the phones location inside tunnels and stuff
>>
>>
>> Unfortunately not. As already pointed out on that list, the
>> accelerometers error would add up itself.
> 
> 
> But if it's as good as GPS for 10-15 sec it should be possible to 
> interpolate 10 - 15 GPS measurements while on the move. Must move fast 
> enough so the start vector can be known. But while moving so fast the 
> accelerometers can be calibrated over time by GPS data cramming 
> absolutely maximum accuracy from them! Maybe detect moving in and out of 
> GPS echoes and compensate the results to.

One 3 axis accelerometer - tied down in a car-holder, with the best of 
the $5-10 accelerometers, gives interesting navigation possibilities.
You can basically use the fact that you know that where the car is 
pointing is where it's going, and that any lateral accelerations are turns.
(this breaks down with banked roads, or drifting the vehicle).
But an error of around a milligee gives only an error of around 5m after 
30s, which isn't at all useless. (around half a kilometer after 5 min)

However, the tiny baseline of the phone means that the differential 
accelerations are tiny.

In the non-tied-down case, where you can't tell anything about the 
phones orientation, you can do almost no navigation.

http://wiki.openmoko.org/wiki/Accelerometer_Fundamentals

The differential accelerations are so small that with the best availble 
accelerometers, you cannot tell the important parameter - roll about 
'down' - the noise makes the roll reading spin randomly round and round 
every few seconds.




More information about the community mailing list