There are few key steps.
        1. Select a location for the antennae (empty hex will maximize rewards)
        2. Mount the antennae with a clear sky view
        3. Connect the station to Internet
        4. Add the miner and your crypto-wallet to the GEODNET console

RTK is so accurate because it removes the primary errors in the GPS/GNSS signals. These errors include localized ionosphere and troposphere induced errors (“space weather”), orbit, and clock errors.

A few nice explanation is found here:
Navipedia
Wikipedia

GPS refers to the United States’ constellation of PNT (Position Navigation and Timing) satellites. GNSS refers to all constellations including EU (Galileo), Russia (Glonass), China (Beidou), Japan (QZSS), and India (Navic).

Space Weather describes the variations in the space environment between the sun and Earth. In particular, Space Weather describes the phenomena that impact systems and technologies in orbit and on Earth. One of those impacted systems is GNSS. Space weather impacts GNSS receivers in several ways. GNSS radio signals travel from the satellite to the receiver on the ground, passing through the Earth’s ionosphere. The charged plasma of the ionosphere bends the path of the GNSS radio signal similar to the way a lens bends the path of light.

PPP makes use of a network of reference stations in order to compute precise estimates of GNSS satellites orbits and clock errors. Nevertheless, it requires fewer reference stations globally distributed as compared with classic differential approaches (e.g. Real Time Kinematics, RTK), and one set of precise orbit and clock data (computed by a processing center) is valid for all users everywhere. However, the ionospheric or Space Weather compensation is less accurate, resulting in a longer and more fragile precise fix.

PPP-RTK combines the techniques of PPP with a denser regional GNSS base stations to produce a better ionospheric (Space Weather) model. PPP-RTK features faster convergence time as compared to PPP. PPP-RTK is also referred to as SSR (State Space Representation) corrections, as like PPP the corrections are computed server side.

A couple of good articles:
https://www.septentrio.com/en/learn-more/insights/gnss-corrections-demystified
https://www.geopp.de/ssr-vs-osr/

Low-Earth Orbit (LEO) satellites offer a promising future for precise positioning. As they are closer to the planet Earth, the signal strength will be higher which will reduced multipath errors as well as the vulnerability to GNSS jamming. LEO satellites also require ground-based reference stations, and GEODNET is investigating how to mine existing and future LEO satellites for valuable PNT data.

While LiDAR and Cameras most fundamental use in autonomous systems is for Perception, they also play a key role in Localization. The sensors are complimentary to RTK and high-precision GNSS positioning. LiDAR and Cameras are frequently used for relative positioning or with the help of an HD Map, can be used for absolute positioning in the map frame. Inertial Measurement Units (IMU’s) provide position based on the numerical integration angular rate and accelerations. IMU’s are useful because they do not have any external dependency; however, IMU’s are susceptible to drift.

GPS, GNSS and even RTK struggles greatly indoors due weak signals and large amounts of multi-path errors from the building walls. Your mobile phone usings WiFi maps and other tricks to improve the accuracy of your indoor location, but it is still not very good. GEODNET is exploring ways to use #DePIN to improve indoor position accuracy.