Development and Implementation of Key Functionalities for a Lunar Rover Guidance Navigation and Control System

In the renewed era of lunar exploration, international collaborations among space agencies and industry are shaping a shared vision of a sustainable human presence on the Moon. Within this framework, robotics and autonomous systems play a crucial role in enabling surface operations, scientific activities, and infrastructure deployment under extreme and communication-limited conditions. The Moon Universal Locomotion System (ULS), developed with the support of ASI and led by Thales Alenia Space Italia (TAS-I), represents a modular and scalable rover platform designed to provide flexible mobility solutions for future planetary missions. This doctoral research, conducted within the ULS initiative, focuses on the design and validation of GNC algorithms and architectures to enhance rover autonomy in perception, mapping, and maneuvering. The work is structured around two main technological axes, complemented by a third, independent activity on LiDAR-inertial system evaluation. The first research stream addresses terrain mapping and traversability estimation. A lightweight, computationally efficient geometric-based mapping pipeline was developed for unstructured environments, enabling real-time terrain assessment through elevation data to estimate slopes and infer traversability. Implemented and validated in a ROS 2 environment, the method was compared to a conventional normal-based approach, showing significantly improved computational performance while maintaining comparable accuracy, an essential advantage given the processing and energy constraints of space missions. To enhance robustness, the study integrates learning-based frameworks that incorporate semantic information. Conducted at the DLR-RMC, this component integrates a neural network for semantic segmentation, trained on the RUGD dataset, into the mapping system of Lightweight Rover Units (LRUs). The resulting geometric-semantic framework combines local geometry and terrain classification to generate adaptive traversal costs, improving environmental understanding and enabling more reliable, mission-aware trajectory planning. The second research stream focuses on autonomous maneuvering for a non-holonomic lunar rover prototype featuring four independently steerable wheels. This configuration supports multiple steering modes, such as crabbing, pure rotation, and Ackermann steering, providing high flexibility for navigation and mission-specific operations, including solar panel alignment and stability on uneven terrain. To exploit this flexibility autonomously, the work introduces an extended ROS 2-based Nav2 architecture, leveraging a behavior-tree approach that combines path planning, dynamic maneuver selection, and trajectory tracking. Validated through extensive simulations and field tests using the European Moon Rover System (EMRS) prototype at the TAS-I RoXY facility in Turin, the system demonstrated robust and context-aware autonomous navigation in lunar-like conditions. The third research activity develops a comparative framework for LiDAR-SLAM systems, assessing performance through metrics such as Absolute Trajectory Error (ATE), uncertainty, feature tracking, and computational cost. Experiments on the MORPHEUS rover at the University of Padova compared two LiDAR-inertial setups: a motorized Ouster OS1 (360° FoV) and a MEMS-based Livox Horizon (limited, non-repetitive FoV), both processed with the FAST-LIO 2 algorithm. Despite differing sensing characteristics, both achieved comparable localization accuracy, demonstrating that MEMS-based sensors can be effective low-cost alternatives to full-view LiDARs. Overall, this thesis advances autonomous navigation for lunar rovers by integrating efficient geometric-semantic mapping, adaptive maneuvering, and experimental validation on multiple robotic platforms. The results contribute to developing more capable, resilient, and resource-efficient autonomy solutions for the next generation of lunar exploration missions.