Improving position estimation in mobile robots based on sensor fusion techniques

In recent years, autonomous systems, robotic technologies, and smart sensor platforms have found widespread application across various sectors. In critical domains such as defense, transportation, healthcare, and agriculture, ensuring that these systems can navigate safely and accurately with environmental awareness has become a fundamental necessity. To address this need, one of the core algorithmic approaches developed is Simultaneous Localization and Mapping (SLAM). SLAM enables a robot or autonomous vehicle to construct a map of an unknown environment while simultaneously estimating its own position. Thanks to advances in sensor technologies, localization systems are no longer dependent on a single data source. Instead, multi-sensor fusion techniques combining stereo cameras, inertial measurement units (IMU), LiDAR, and odometry are employed. While this integration significantly improves the accuracy and robustness of localization, it also introduces technical challenges such as time synchronization, calibration inconsistencies, data integrity issues, and sensor noise. Therefore, in addition to classical SLAM algorithms, complementary methods aimed at improving localization accuracy have become increasingly important. xvii In this study, position estimates obtained from a SLAM system utilizing multi-sensor data were enhanced using artificial intelligence models such as LSTM and MLP. The performance of these correction models was evaluated by comparing their outputs with ground truth data using both quantitative metrics (e.g., RMSE) and visual analyses. Furthermore, the effects of data preprocessing, sensor types, and synchronization structures on localization accuracy were examined in detail. The findings demonstrate that AI-assisted localization systems offer significant improvements compared to classical SLAM-based approaches.