Mapping and position estimation in indoor environments without Global Navigation Satellite Systems

Position estimation and mapping in the absence of Global Navigation Satellite Systems (GNSS) is an important and challenging problem. The problem of position estimation in the absence of GNSS has been studied for many years. Especially with the increase in processing power, ergonomics, and availability of cameras and inertial measurement units (IMUs), the idea of using them together to solve position estimation problems has grown considerably. This thesis discusses the mapping and position estimation processes of robotic systems in indoor environments where GNSS is absent or ineffective. Nowadays, providing accurate and reliable positioning in indoor environments where GNSS signals are blocked or inaccessible is of critical importance, especially for the defense industry. In this context, visual-inertial simultaneous localization and mapping (ViSLAM) algorithms provide an effective approach for indoor positioning. In this thesis, the camera system is integrated with an IMU and visual and inertial data are combined. The algorithm used performs mapping and position estimation by simultaneously processing visual data from dual (stereo) cameras and IMU data. To improve the accuracy and performance of the algorithm, detailed IMU calibration has been performed. This ensures that the system can accurately process the sensor data. The main advantages of the ViSLAM algorithms, namely noise robustness and high precision, are analyzed in detail in this study. In the experimental platform used in the study, both the image transmitted by the stereo camera and the data from the inertial measurement unit sensor were processed using the Robot Operating System (ROS) infrastructure and integrated with the ORB-SLAM3 algorithm. The developed system enabled a human/robot moving in an indoor environment to accurately map its environment and determine its own position with high accuracy. This thesis demonstrates the potential of using the ViSLAM algorithm with the Luxonis Oak-D-Pro camera in real indoor applications and contributes to the literature in this area. The fact that the Oak-D-Pro camera has not been used in the literature makes this study different from others. The results obtained show that the system developed with the Oak-D-Pro camera is a promising solution for both academic and industrial applications.