Development of Localisation System Framework and Ground Truth Generation Pipeline for KuupKulgur

Abstract

This thesis presents the development of a localisation system framework that includes a custombuilt Unscented Kalman Filter (UKF) library for lunar rovers and a lidar-based ground truth generation method for evaluating its performance. The work is motivated by the Estonian microrover project KuupKulgur, which requires a localisation system as part of its larger autonomy software development. The created system needs to be modular to support the project’s changing requirements due to its early stage of development, which includes different robot mechanical modifications and sensor configurations. A C++ localisation library that includes the UKF was developed together with a Robot Operating System (ROS) package that enables future integration with the existing software stack of KuupKulgur. The library supports modular state and measurement vector handling with easily interchangeable system models. For ground truth generation and evaluation, a ROS package was implemented that utilises the Direct LiDARInertial Odometry (DLIO) algorithm with the Ouster OS-1 lidar. This work serves as the basis for estimating lunar rover localisation and verifying the results. Verification is performed by comparing localisation results to ground truth generated from a lunar analogue site using the microrover KuupKulgur.