Abstract
Improved accuracy and reliability of the user coordinates is expected if GNSS data is processed using the network- RTK technique. Some implementations of conventional network-RTK require bi-directional communications such as GSM/GPRS since the network server transmits network-corrections to a rover, while the rover is required to transmit its (approximate) coordinate information to the server. The data transmission from the rover is kept to a minimum because the server needs the rover’s coordinate information only for “initialization” or “handover”. On the other hand, a server-based network-RTK requires raw measurements from the rover receiver in order to computer the rover’s coordinates. There are several commercially available network-RTK systems, however, server-based techniques are relatively new in surveying. In GNSS geodesy, server-based network processing has always been the norm, though rarely in real-time. However, with the move of the IGS towards a “real-time future”, the distinctions between server-based network- RTK implementations for survey users and for geodesy will be blurred. This paper describes a server-based network-RTK design that computes a rover’s coordinates in the required reference system by taking advantage of existing GNSS reference network infrastructure, instead of broadcasting corrections or data to users and placing the onus of obtaining a final solution on clients and their equipment. The system architecture of the proposed server-based network-RTK is described in the paper. The basic assumption is that distributed-computing is necessary to cope with simultaneous requests from hundreds of clients. Distributed-computing allows computers to efficiently communicate and individually process data, which is different from networked-computing. The proposed design features receiver-independency (and therefore interoperability), network-scalability, and userflexibility. Test results show that the proposed serverbased network-RTK provides centimetre-level positioning for regional GPS networks with medium baseline lengths (20-35km) and medium-to-long baseline lengths (50- 100km).