Previous Mars and Moon rover missions have highlighted a limitation in mobility, with surface rovers traversing only a few tens of meters per day, at speeds in the order of 10 cm/s. This limitation can be primarily attributed to the rover locomotion system from one side, and on the other by the lack/limited on-board autonomous capabilities.To address these limitations, ESA issued a series of activities on fast autonomous rovers. GMV is leading the RAPID and FASTNAV projects, aiming at developing a semi-autonomous rover capable of safely traversing lunar terrains at high speeds (around 1 m/s), using a visual navigation based Guidance, Navigation and Control (GNC) system. The RAPID project was the foundation of this work, in which the rover platform was designed, manufactured, and a first version of the on-board software was developed. The follow-up FASTNAV project builds upon the work of RAPID to establish an innovative multi-mode continuous drive GNC system. This new architecture aims to achieve an improved autonomy and faster speeds, with different guidance modes selected according to the terrain type, using a combination of techniques (AI and classical computer vision), while also improving the existing GNC algorithms.This paper shall present the experience obtained during the development of the rover platform and the GNC system across both activities. It shall provide an overview of the challenges and risks, the decisions made, a detailed technical overview of the GNC architecture and components, and the results observed during the analogue field tests.

Terrain classification is crucial for the successful execution of autonomous navigation and path planning during Mars rover missions. This study focuses on enhancing the rover's capability to traverse the Martian surface by investigating the integration of advanced semantic segmentation models based on deep learning. The aim is to identify the most effective deep learning model from recent advancements and establish efficient training approaches.The study selected the state-of-the-art U-Net and DeepLabV3+ models for further assessment and evaluation, utilizing both the AI4Mars and ESA's LabelMars datasets. Techniques such as preprocessing, augmentation, and various loss functions were investigated to improve model performance and class imbalance issues are tackled. To mitigate overfitting, regularization techniques like weight decay and early stopping were applied, ensuring robust model training. Additionally, to further enhance the model’s performance, especially in recognizing rare classes, the study investigated the use of state-of-the-art GAN models for generating new images and expand training sets.Our findings reveal that excluding the background class from training and testing significantly improves model performance. Using early stopping regularization reduces the training time drastically while giving high model performance. Notably, the DeepLabV3+ model surpasses the performance reported in existing literature, achieving a maximum segmentation accuracy and Mean Intersection over Union (mIoU) of 99% and 87% on the AI4Mars dataset, and 87% and 72% on the LabelMars dataset, respectively. The integration of GAN-generated images into training further improved rare class performance by up to 2%. These advancements in deep learning models for terrain classification promise to significantly enhance the capabilities of Mars rovers in autonomous navigation and path planning.

The future assembly of the Lunar Orbital Platform Gateway, on a near 9:2 resonant Rectilinear Halo Orbit, requires the necessity of designing safe and reliable strategies to perform rendezvous and docking with the station. The paper describes and tests a strategy to guarantee the safety of an entire rendezvous manoeuvre in the proximity of the Moon L2 Lagrangian point with respect to a specific set of failures. The safety in the far range is guaranteed through the automatic allocation of selected hold-points, whereas during close range rendezvous the safety is actively guaranteed in the presence of selected failures. The main goal of the paper is to contribute to the very limited literature about autonomous design of the guidance for rendezvous in presence of a non negligible third body influence, where safety considerations are paramount. The proposed approach is based on the exploitation of the manifold theory used in the circular restricted three body problem model to guarantee the passive safety in the far-range section, and on an optimal and reliable active collision avoidance manoeuvre to guarantee the safety for the close range approach.

This paper describes key technologies for the use of the six-legged walking system Mantis in a multi robot team performing cooperative tasks for the construction of an In-Situ Resource Utilization (ISRU) facility on the Moon. Autonomous multi robot cooperation is one of several key technologies that hold promise for In-Situ space exploration and ISRU facility construction. Therefore, the PRO-ACT (Planetary RObots deployed for Assembly and Construction Tasks) project aimed to develop a multi robot team which can act (semi-) autonomous. PRO-ACT is a European project with many partners working on space robotics technologies, the so-called building blocks. These building blocks have been implemented and extended on Mantis to cover the scope of multi-robot cooperative scenarios. To allow the different partners to execute commands without considering the robot's control framework, a communication interface was developed to provide a common and generic way to send commands and receive sensor data. It facilitates access to all robots and to their simulated counterparts. Due to the impact of COVID-19, most of the testing, including the final demonstration, was performed remotely, with robots available at the partners' premises.

ESA is working with NASA to plan and carry out an international Mars Samples Return (MSR) campaign between 2020 and 2030. A relevant part of the upcoming MSR mission is the Sample Fetch Rover (SFR), tasked to collect sample tubes of Martian soil prepared by Mars2020 rover Perseverance. This work focuses on the localization capabilities of SFR and the potential reuse of the functionalities present in the ExoMars rover. Visual Odometry (VO), a vision-based localization algorithm, is often the main component of the localization process in planetary robotics. The goal of this study is to investigate the possibility of transferring the ExoMars VO solution to a valid SFR implementation, compliant with mission requirements. First, the main differences between the two missions, SFR and ExoMars have been studied, in order to identify the most critical parameters for the VO process. Then, using a testing rover available in the Planetary Robotics Lab (PRL) at European Space TEchnology and research Centre (ESTEC), the effect of the previously identified parameters on the VO performances was evaluated, identifying the most crucial ones and proposing some solutions to face them. This work could lead the way to future studies about the localization for the Sample Fetch Rover and what are the main and most critical factors that would have to be taken into account in order to achieve an accurate and reliable localization system.