Coordinate System Alignment and Sensor Fusion for Multimodal Localization System

Recently, the development of new autonomous mobility solutions such as robots and AGVs has been actively pursued, which significantly contribute to labor reduction and efficient logistics. Autonomous driving requires a broad range of technologies, including localization, navigation, and fleet management. Among these, high accuracy localization is especially important to achieve a stable system. It is common to integrate multiple localization technologies, but these technologies often have their own coordinate systems, making it necessary to accurately align them prior to system integration. This paper reports on a method of achieving precise alignment of different coordinate systems by using multiple reference anchor points to statistically derive their relationships in the shape of a coordinate transformation matrix. Additionally, we enlarge the study to ensure alignment accuracy for coordinate systems with non-linear distortion, which is often seen in distorted maps made by SLAM in environments with high error accumulation. The effectiveness of the method was tested by measuring localization accuracy on an outdoor road that was more than 1/4 mile long. Localization error due to coordinate system misalignment caused by map distortion was effectively suppressed, resulting in stable localization throughout the testing route. Furthermore, an additional experiment was conducted to test indoor/outdoor seamless localization. The coordinate system alignment method was applied to accurately align GNSS and SLAM, enabling stable sensor fusion for a multimodal localization system. By using this method, it is possible to align coordinate systems when integrating various localization technologies, such as LiDAR-SLAM or Visual-SLAM with unique map coordinates, GNSS with UTM coordinates, and beacons or AR tags with unique coordinate systems.