The objective of this thesis is to develop, test and analyze the performance of a high accuracy GNSS PPP-RTK (Precise Point Positioning- Real Time Kinematic) operational framework based on a dense network of peer users, through the exchange of precise Ionospheric and Tropospheric estimates between peers. The PPP-AR (PPP Ambiguity Resolution) technique, based on the use of precise satellite orbit and clocks, needs satellite biases estimates provided to receivers by external sources. It also usually requires a long convergence time for fixing carrier phase ambiguities to their correct integer values. The novel approach developed herein reduces convergence time in a network of closely spaced (on the order of a few km or less) user receivers performing PPP-RTK with the addition of sharing precise STEC (Slant Total Electron Content) and ZTD (Zenith Tropospheric Delay) estimates. These are applied by the users as constraints for faster convergence to sub-decimetre accuracies. This approach is named Peer PPP-RTK. Peer solutions are processed through a decentralized Federated Kalman Filter approach, where all peers are considered as independent local filters, deriving autonomously their own federated solutions and integrating closer peer ionospheric and tropospheric estimations. The resulting processing engine is tested first through simulations and then in a real GNSS network of permanent reference stations densified through VRS (Virtual Reference Station) representing kinematic users. A performance analysis to assess the impact of network receiver spacing with different scales of ad-hoc networks is carried out in order to simulate possible application scenarios (e.g. cadastral surveying, car information sharing and crowdsourcing in a Collaborative – Intelligent Transportation System perspective). To evaluate the performance of a dense network of peers in a city, a Monte Carlo simulation over 500 randomly distributed networks and different application scenarios is carried out and convergence time statistics are derived through Minimum Spanning Tree analysis. Analysis of these tests shows a percentage convergence time improvement of the Peer PPP-RTK approach with respect to the PPP-AR case for a network of 50 close peers in an area of 100 km2 of 37% (in harsh scenarios with high shadowing and relevant convergence time increase) to 43% (in nominal scenario with nominal PPP-AR convergence times), while preserving an accuracy of at least 5 cm after convergence. This is a highly significant improvement as it has the potential to result in minor operating costs for scores of applications. The PPP-RTK collaborative approach paves the way for the development of a Global PPP-RTK development for a host of large scale applications while reducing the number of Reference Stations required and associated infrastructure costs.