Go Down

Topic: Optimization of GPS accuracy - Chronometer (Read 952 times) previous topic - next topic


Aug 23, 2015, 08:14 pm Last Edit: Aug 23, 2015, 08:52 pm by 007good

I just got an Adafruit Ultimate V3 GPS.
It deliver "only" DDMM.MMMM and says to be 1.8m accurate - 10Hz. https://learn.adafruit.com/downloads/pdf/adafruit-ultimate-gps.pdf (page 4 for spec).

I know I passed the finish line when the segment of point N-1 to N cut the specified finished line. Because of 10Hz, I may be a maximum of 0.1sec late, but as I know the position of the N point, I know the distance from the finish line to the point N, as I also know the speed, I can perfectly know when I passed the line.

Without considering the speed accuracy (which is 0,1m/s considering the spec), the position accuracy cause a 1,8meter uncertainty. Which lead to an uncertainty of 0,06sec (at 100 km/h) on the lap time.

That is not bad, but are my calculation correct ? Can we improve it ?


Is the 1.8m CEP, RMS, 2DRMS or R95?

For example, if 1.8m is CEP, then 50% of the time, your positional accuracy is <3.6m along-track error.

Can of worms.


According to the datasheet of the GPS which i just checked (MTK3329), it says 3m 2D-RMS.. So I must stick to 3m instead of 1,8m I guess, so around 0,1sec (at 100km/h) uncertainity, this begin to be a lot :(


So I must stick to 3m instead of 1,8m I guess
That is the best you can expect. Often, it can be much worse.



So... anyone knows how to improve accuracy ? thanks to the alorithm probably ?

industrial gps chronometer claim to be 0.001 sec accurate (and sell the finish product around 200$), so i guess they don't have super accurate gps either... how they do it ?


industrial gps chronometer claim to be 0.001 sec accurate
That's the time accuracy, NOT a position rate of 1000Hz.  A chronometer is for "measuring time",  so that's just an accurate clock.  A ublox 6M can know the time within microseconds when it has a location fix within 300m, and within nanoseconds when it has a location fix within 1m, so I'm not sure why it's $200.

There's no magic algorithm for increasing accuracy, contrary to what you see in movies.  You can average your location over a period of time to decrease the error, but that's not very helpful for a race.  :P

Also, the speed and altitude have their own errors, so you don't necessarily "know the speed".  Sensing the wheel rotation would be much more accurate... except if there's slippage.

You could add an Inertial Measurement and/or a compass to decrease the error, but that software is much more complicated, like that found in drones.

If you need to know when you pass a line, it's much easier to put a stationary sensor at the line: an IR or laser beam, a bumper or string switch, a magnet+hall effect switch, or a switch plate you run over.

Really, I used to be /dev.  :(

Go Up