A Study on Improving Performance of Network RTK through Tropospheric Modeling for Land Vehicle Applications

Abstract

Network Real-Time Kinematic (RTK) has been developed in the late 2000s to overcome the limitation of the conventional single baseline RTK. It is capable of achieving cm-level positioning accuracy while reducing the number of reference stations required to cover the same amount of area compared to that of the RTK. However, Network RTK has been widely used for mostly static applications such as surveying and geodesy. Recently, many researchers have been studying on application of the Network RTK for dynamic purposes and users, especially for land vehicles, such as automatic vehicles, smart car systems, traffic control and monitoring vehicles carrying hazardous material, due to the increasing demands for user convenience and safety. This thesis focuses on improving the performance of Network RTK through modeling tropospheric delay for land vehicle applications. The post-processing Network RTK system was developed by the GNSS laboratory at Seoul National University in 2008. This thesis further strives to generate robust and accurate corrections by improving the estimation performance of the integer ambiguities between reference stations that constitute a network by estimating tropospheric wet vertical delay and multipath errors for medium-baselines. In addition, the technique for adjusting integer ambiguity levels among networks is proposed to enable dynamic users to continuously achieve high-accuracy positioning regardless of network switching. Furthermore, a new multiple corrections modeling method is proposed to improve user accuracy. According to how it generates corrections, Network RTK is classified into three techniques: VRS, MAC and FKP. This thesis generates the MAC-based Network RTK corrections since the other methods are known to be deduced from the MAC approach. In order to generate MAC correction, precisely estimated integer ambiguities between reference stations are necessary. The baseline length of the two reference stations in a network is typically 50 to 70km. Therefore the conventional single baseline RTK cannot be used to estimate those ambiguities. This thesis utilizes Kalman filters to estimate tropospheric wet zenith delay and multipath errors for accurate estimation of such integer ambiguities. The dynamic users can receive different corrections from different networks because they are in constant motion and therefore the network from which users receive correction can be switched. Regardless of network change, the user should be able to continuously calculate accurate positions at any location. In order to fulfill this requirement for the land vehicles, the integer level adjustment technique is proposed. Lastly, users should be able to combine the multiple corrections received from reference stations in a network for their location to eliminate their GPS errors. Although many researchers have developed various correction modeling methods, the method that this thesis proposes is considered to be new as it considers the physical characteristics of tropospheric delay over height. In order to evaluate the performances of the implemented and proposed algorithms, the following tests are conducted: First, the estimation performance of the medium-baseline tropospheric wet zenith delay, multipath errors and integer ambiguities are evaluated using both simulation and real GPS measurements. Second, the ambiguity level adjustment technique is proposed and verified through simulation for Networks distributed all over the South Korea. In addition, dynamic tests are conducted to evaluate the performance of the generated corrections of the MAC-based Network RTK and user positioning accuracy before and after the ambiguity level adjustment. Lastly, the performance of the proposed correction modeling method is evaluated using both the simulated and real GPS measurements for Networks in the USA.