چكيده به لاتين
In the modern and advanced world today, precision in positioning and orientation plays a vital role in various applications such as robotics, vehicles, and self-driving cars. One of the major challenges in positioning systems is excessive reliance on GPS signals, which may have lower accuracy and reliability under specific conditions. Additionally, combining GPS with Inertial Navigation System (INS) alone cannot provide a stable solution against cumulative errors and can only tolerate GPS signal interruptions for a few seconds. This thesis begins by examining existing positioning systems and their improvement methods. Subsequently, it introduces an innovative strategy for integrating imagery with INS, providing a significant performance improvement compared to INS alone. The research also presents an advanced approach to enhance navigation and positioning accuracy by combining GPS data with Visual-Inertial Odometry (VIO) using graph decomposition and intelligent algorithms. The proposed system is implemented and tested on the Jetson platform, and the comprehensive results of these experiments are presented in this thesis. Using the KITTI dataset, the GPS/VIO system demonstrated a remarkable improvement of 84.59% and 88.806% in GPS trajectories' accuracy at 10 and 27 seconds, respectively. This GPS/VIO system, based on factor graph and intelligent methods, is suggested as an effective solution to address concerns related to accuracy and reliability in GPS-based navigation. Furthermore, data preprocessing techniques have been employed to enhance accuracy, including optimizing image data by adjusting contrast and brightness levels, as well as reducing noise in IMU data. This led to a significant 44.7% improvement in the accuracy of predefined route paths, detailed in the results chapter.
