Integration of Navigation Sensors Based on Factor Graph Optimization

Global Navigation Satellite Systems (GNSS) and Inertial Navigation Systems (INS) play pivotal roles in modern navigation and positioning applications. The integration of these two technologies has become crucial for achieving high-accuracy and robust navigation solutions in various domains, such as autonomous vehicles, aviation and mobile devices. These two technologies are characterized by complementary features, with GNSS providing absolute position but suffering from dependency on external conditions and INS offering continuous, high-rate relative motion data but being affected by error accumulation over time. Their integration aims to emphasize strengths while minimizing weaknesses for robust and accurate navigation solutions. Several integration approaches, of different complexity, can be found in literature. In this thesis a factor graph framework has been adopted to perform such integration. Factor graphs have proven as a powerful mathematical framework for modeling and solving complex estimation and optimization problems as Simultaneous Localization and Mapping (SLAM). Recently, because of their flexibility, factor graphs have emerged as an alternative method for GNSS positioning. Factor graphs describe positioning problems in terms of optimization problems, allowing the solution to be obtained over multiple iterations, differently from other traditional navigation filters such as Extended Kalman Filter (EKF). This means that the involved functions are linearized multiple times and time correlation between consecutive epochs is better exploited. Moreover, positions related to previous instants can be kept inside the graph so that, as new positions are estimated, the gained information can be used to refine the old solutions, obtaining a better performance in post-processing. These features, along with many robust estimation techniques that have been developed for factor graphs, allow to obtain positioning solutions which can be more robust in challenging environments, such as urban scenarios. This thesis provides a comprehensive overview of GNSS and INS technologies as well as the mathematical formulation of factor graphs. The framework of a GNSS/INS tight integration based on factor graphs is then developed. Finally, an analysis of the results obtained deploying this formulation on a MATLAB receiver is performed and compared to the solution obtained from an EKF. The data-set used for this purpose is taken from a urban scenario (città di Torino). In conclusion, the results obtained in this thesis prove the advantages coming from the integration of GNSS and INS in a factor graph framework, offering increased accuracy and robustness in complex environments.