Acquisition and parsing of GPS data

alphaville commented 7 months ago

Describe the solution you'd like

I have uploaded the data we collected on 28/11/23 https://gist.github.com/alphaville/995a15ef74e8d5b3128c87667b5bfdaf. This is how the data looks like

$GPGLL,5434.81291,N,00556.15593,W,145737.00,A,A*70
$GPRMC,145738.00,A,5434.81285,N,00556.15627,W,0.351,,281123,,,A*63
$GPVTG,,T,,M,0.351,N,0.650,K,A*27
$GPGGA,145738.00,5434.81285,N,00556.15627,W,1,05,1.28,64.4,M,52.5,M,,*79
$GPGSA,A,3,12,32,24,15,19,,,,,,,,2.51,1.28,2.16*02
$GPGSV,3,1,10,02,02,344,37,12,56,202,22,15,26,166,27,17,,,35*49
$GPGSV,3,2,10,19,34,071,36,22,11,050,32,23,18,236,,24,72,108,35*72
$GPGSV,3,3,10,25,21,223,,32,26,309,41*75

Documentation

The data are structured according to the NMEA Standard Logs protocol. Examples:

GPGLL - Geographic Position
GPRMC - GPS specific information incl. latitude/longitude and speed
GPVTG - Track Made Good and Ground Speed
GPGGA - GPS fix data and undulation
GPGSA - GPS DOP and active satellites
GPGSV - GPS Satellites in View

First of all we need to decipher what all these mean and how the data are structured (read the docs).

Issues to address

We interface the GNSS over the serial at a baud rate of 500,000. We can write a simple interface that reads the data from serial (using pyserial) and parses it
We should follow what we did here for the anemometer: a thread runs in the background, collects data and stores it in memory, and then we can access is with a non-blocking function (or get a moving average, median, etc).
For some reason the data seems to be coming in at 1Hz, which is too slow. Can we speed this up?

~Accuracy~

~In terms of absolute accuracy, it seems to be quite bad...~ (Update: My bad! See my comment below)

~There seems to be negligible variance when the antenna is perfectly static.~ (Update: It depends... see below)

We should see how sensitive it is: we should move it by 1m and see what measurements we get. (TODO!)

alphaville commented 7 months ago

@pdavid747 @Runway27 Note that if you write your own code you won't have to use gpsd at all. Actually you shouldn't use it because it won't let you read the serial. Start by reading the documentation to understand what sort of data we get from the GPS and then build a simple DTO: a Python class whose attributes are the quantities the GNSS measures.

alphaville commented 7 months ago

@pdavid747 @Runway27 @jamie-54 u-center is a software provided by u-blox and should allow us to configure the GNSS module and increase the update rate

alphaville commented 7 months ago

Accuracy of GPS measurements and conversion to degrees

⚠️ The GPS coordinates are in degrees-minutes-seconds; after the conversion to a decimal representation, it turns out that the coordinates are fairly accurate.

@pdavid747 use the following function to convert the GPGLL data to degrees:

def deg_min_sec_to_decimal(dms, dir="N"):
    """
    Convert degrees-minutes-seconds from GPGLL to degrees (deciman)

    :param dms: degrees-minutes-seconds (must be positive)
    :param dir: direction (can be one of 'N', 'W', 'E', 'S')
    """
    dms_abs = np.abs(dms)
    degrees = np.floor(dms_abs/100)
    minutes = np.floor(dms_abs - 100 * np.floor(dms_abs/100))
    seconds = 10 * (dms_abs - np.floor(dms_abs))
    deg_decimal_abs = degrees + minutes / 60 + seconds / 3600
    dir_sign = -1 if (dir in ["S", "W"]) else 1
    return dir_sign * deg_decimal_abs

alphaville commented 7 months ago

I installed PyGPSClient on Mac:

It allows to set the measurement update rate and many more settings.

alphaville commented 7 months ago

Scatter plot of estimated positions while static:

while being connected to 3-8 satellites (the antenna is placed on metallic plate, but indoors having a decent but not perfect view of the sky; it has an elevation of around 75 degrees). We seem to be getting an accuracy of about ±10m at best, but we might get better results outdoors.

Some more results...