From OpenStreetMap Wiki
Jump to: navigation, search

I bought my first GPS (GARMIN GPS2+) in 1998 for a friend with a sailing boat, but before I could hand it over I became so fascinated that I bought another one for myself. Now it died after many faithful years of service and I replaced it by a GPSmap 60csx, which I like a lot. Recently I discovered OSM and try to help by mapping:

- details in Hannover-Nordstadt and Hannover-Herrenhausen

- street network of Havelse near Hannover

- some ski slopes in Austria and a village in South Germany

- ... more to come as I have time

I am also doing some experiments to improve the GPS precision by kinematic relative positioning.

By using two receivers (a fixed "base" with good sky view and a "rover"), recording and post-processing the raw data (pseudorange, carrier phase and Doppler), some errors can be eliminated, namely the satellite clock errors and the ionospheric and tropospheric propagation delays.

Furthermore, postprocessing allows better filtering of the data ("backward RTS smoother") than the forward-only Kalman filter to which real-time devices are restricted.

So far, I have already overcome some big initial obstacles:

- recording the data, writing and reading RINEX files.

- satellite orbit interpolation from SP3 files.

- coordinate conversions.

- clock jump correction.

- adaptive weighting of the raw measurements with a model based on elevation and SNR.

- forward Kalman filtering and backward RTS smoothing using pseudoranges and Doppler velocities.

The results are already somewhat better than what my GARMIN produces (this is of course no fair comparison, since I have a base station and allow postprocessing).

Further improvements are expected by using the carrier phase. The main problem is to detect and possibly correct cycle slips.

Integer ambiguity resolution seems almost impossible with a moving single-frequency receiver, but some smoothing should be possible and helpful.

So far the precision is apparently limited by multipath, no surprise. Ground planes and choke rings seem impractical and expensive for bicycle-based OSM mappers, but when I have time I may try some other antennae.

For reliable tests of the achieved precision, I am lacking some well-defined reference points with precisely known coordinates. Now I judge performance mainly by the noise level and repeatability.