I've read a number of articles on RTK GPS. I'm only familiar with using RTK GPS with an RTK Base Station, and I understand how the corrections can be broadcast to a rover. When would you use RTK GPS without using a base station, and is this practical?
In all cases, L-code receivers need to resolve the "integer ambiguity" between signals received from the observable satellites in order to achieve their differential GPS accuracies.
So, I believe there are two questions you may be asking: 1) How can I achieve differential GPS accuracy using a single roving L-band receiver? or 2) How can I get the most out of an RTK receiver in autonomous mode?
For question 1: Alternatives to a "traditional" base station include subscribing to a virtual reference station network. -- these subscription services deliver real-time correction parameters over cellular-IP addresses.
- the corrections are "smeared" (read triangulated, least-squares averaged) over a mesh of correction base stations
- uploaded to the internet then made available for download.
Receivers download them and apply the differential corrections on-the-fly during RTK observations just as though they were receiving a signal from a dedicated base on a known point.
Question alternative 2 is less beneficial as autonomous L-band positioning is only slightly more accurate than C-code autonomous positioning. The better option to get full value from a single L-band receiver is PPK (post-processed Kinematic) where multiple autonomous surveys can be bundle-adjusted together with known baselines to reduce uncertainties and obtain RTK accuracy.
NOTICE that nowhere did I address any issue of "precision" because that is a non-issue when comparing multiple methods using a single device -- the precision is constant (and assumed to be arbitrarily "precise-enough".)
Bottom line, the satellite constellation, atmosphere, and correction baselines will dominate the solution accuracy and the precision is device-dependent.