The increasing importance of Vision-Based Navigation (VBN) algorithms in space missions raises numerous challenges in ensuring their reliability and operational robustness. Sensor faults can lead to inaccurate outputs from navigation algorithms or even complete data processing faults, potentially compromising mission objectives. Artificial Intelligence (AI) offers a powerful solution for detecting such faults, overcoming many of the limitations associated with traditional fault detection methods. However, the primary obstacle to the adoption of AI in this context is the lack of sufficient and representative datasets containing faulty image data. This study addresses these challenges by focusing on an interplanetary exploration mission scenario. A comprehensive analysis of potential fault cases in camera sensors used within the VBN pipeline is presented. The causes and effects of these faults are systematically characterized, including their impact on image quality and navigation algorithm performance. To support this analysis, a simulation framework is introduced to recreate faulty conditions in synthetically generated images, enabling a systematic and controlled reproduction of faulty data. The resulting dataset of fault-injected images provides a valuable tool for training and testing AI-based fault detection algorithms. The final link to the dataset will be added after an embargo period. For peer-reviewers, this private link¹ is available.

Reinforcement learning has proven its power on various occasions. However, its performance is not always guaranteed when system dynamics change. Instead, it largely relies on users’ empirical experience. For reinforcement learning algorithms with an actor–critic structure, the critic neural network reflects the approximation and optimization process in the RL algorithm. Analyzing the performance of the critic neural network helps to understand the mechanism of the algorithm. To support systematic interpretation of such algorithms in dynamic control problems, this work proposes a critic match loss landscape visualization method for online reinforcement learning. The method constructs a loss landscape by projecting recorded critic parameter trajectories onto a low-dimensional linear subspace. The critic match loss is evaluated over the projected parameter grid using fixed reference state samples and temporal-difference targets. This yields a three-dimensional loss surface together with a two-dimensional optimization path that characterizes critic learning behavior. To extend analysis beyond visual inspection, quantitative landscape indices and a normalized system performance index are introduced, enabling structured comparison across different training outcomes. The approach is demonstrated using the Action-Dependent Heuristic Dynamic Programming algorithm on cart–pole and spacecraft attitude control tasks. Comparative analyses across projection methods and training stages reveal distinct landscape characteristics associated with stable convergence and unstable learning. The proposed framework enables both qualitative and quantitative interpretation of critic optimization behavior in online reinforcement learning.

This paper introduces a motion planning method for capture of tumbling objects using a free-floating space robot. The proposed approach incorporates an improved Rapidly Exploring Random Tree Star (RRT*) algorithm enabling obstacle avoidance and generating desired trajectories for the robot's end-effectors. Additionally, a multi-layer optimization process and a greedy policy are proposed to achieve singularity avoidance, joint velocity, and acceleration optimization by leveraging the robot arm's joint energy distribution, torque, and manipulability. By adopting this motion planning strategy, the space robotic system demonstrates improved performance in obstacle and singularity avoidance, without the need for inverse Jacobian matrix calculations. Furthermore, the multi-layer optimization process enhances trajectory smoothness and reduces end-effector vibration through energy and torque optimization. This research contributes to advancing space robotic systems by enhancing the entire energy and torque consumption on motion planning to make the end-effector move smooth and reduce the vibration.

Traditional anomaly detection techniques onboard satellites are based on reliable, yet limited, thresholding mechanisms which are designed to monitor univariate signals and trigger recovery actions according to specific European Cooperation for Space Standardization (ECSS) standards. However, Artificial Intelligence-based Fault Detection, Isolation and Recovery (FDIR) solutions have recently raised with the prospect to overcome the limitations of these standard methods, expanding the range of detectable failures and improving response times. This paper presents a novel approach to detecting stuck values within the Accelerometer and Inertial Measurement Unit of a drone-like spacecraft for the exploration of Small Solar System Bodies (SSSB), leveraging a multi-channel Convolutional Neural Network (CNN) to perform multi-target classification and independently detect faults in the sensors. Significant attention has been dedicated to ensuring the compatibility of the algorithm within the onboard FDIR system, representing a step forward to the in-orbit validation of a technology that remains experimental until its robustness is thoroughly proven. An integration methodology is proposed to enable the network to effectively detect anomalies and trigger recovery actions at the system level. The detection performances and the capability of the algorithm in reaction triggering are evaluated employing a set of custom-defined detection and system metrics, showing the outstanding performances of the algorithm in performing its FDIR task.

Traditional laser communication terminals are limited to point-to-point links, which constrains their scalability and flexibility for global networks that require simultaneous connections with multiple targets. While multiple single-beam terminals can expand capacity, this approach multiplies Size, Weight, Power, and Cost (SWaPC), limiting scalability. Multi-beam laser communication terminals offer a promising alternative, though the design of an effective beam steering system remains a key challenge. This paper explores the design process of such a system, providing an overview of multi-beam steering literature as well as an comparison and trade-off of existing space-borne multi-beam steering technologies. It also analyzes insights from related fields such as terrestrial laser communications, LiFi, single-beam laser communication, optical cross-connects, radio links, and multiple target tracking. Key system functions are identified and visualized in a function flow diagram, and various design options are evaluated, culminating in a design options tree which serves as a design recipe. Two application scenarios, involving high and low target densities, demonstrate that steering systems based on micro-mirror arrays and spatial light modulators present significant advantages over alternatives. This study offers a comprehensive framework for designing multi-beam steering systems for space-based laser communication terminals.

There is an increasing interest in science and exploration missions in the Earth-Moon system. These missions call for simple, efficient and accurate timing onboard of these spacecraft. A heterogeneous onboard timing system is proposed, presented and characterized, which comprises an Oven-Controlled Crystal Oscillator (OCXO) and one or several Chip Scale Atomic Clocks (CSACs), providing nano-second timing accuracy on time scales of seconds to hours. To extend the stability of this system to days, weeks and months, it is assisted by a Global Navigation Satellite Systems (GNSS) timing receiver, which is operated in snapshot mode and activated only for very short time durations of a few seconds on a daily or weekly basis, based on the mission needs. This innovative GNSS assistance replaces the ground-based synchronisation and offers an autonomous operations of the onboard timing system. In this way, a highly accurate yet very low Size, Weight and Power (SWaP) system can be realized, which can be used on very small satellites down to nano-satellites. The onboard architecture of this system is presented and characterised in terms of its achievable Allan deviations as well as its SWaP. It is found that the average power consumption can be as low as 2.2 W, based on a daily GNSS synchronisation schedule, significantly reducing the power requirements over current state-of-the-art solutions. An even less complex variant of this innovative system can be employed for any deep space or planetary mission, replacing the onboard GNSS timing receiver by a traditional time synchronisation mechanism using a terrestrial ground station, or, in the further future, using a pulsar.

Steering multiple laser beams using spatial light modulators (SLMs) creates unwanted diffraction and reflections that are not modulated by the SLM, which can make beam tracking difficult. A novel, to the best of our knowledge, and simple beam steering methodology is proposed, which aims at reducing the influence of this clutter while maintaining tracking performance. The beam(s) are deliberately defocused before steering with a superposition of a phase ramp and Fresnel lens (PRFL) phase screen on the SLM. As a result, the non-modulated reflections and diffracted light are decreased in relative intensity to the steered beam, in turn allowing simple and standard peak intensity and center of gravity (CG) algorithms for tracking. Hardware demonstration shows tracking performance using the PRFL remained on-par with more complex filtering approaches while adding no additional hardware. This method has potential to improve the communication performance of multi-beam laser communication terminals.

There is an increasing interest in science and exploration missions in the Earth-Moon system. These missions call for simple, efficient and accurate timing onboard of these spacecraft. A heterogeneous onboard timing system is proposed, presented and characterized, which comprises an Oven-Controlled Crystal Oscillator (OCXO) and one or several Chip Scale Atomic Clocks (CSACs), providing nano-second timing accuracy on time scales of seconds to hours. To extend the stability of this system to days, weeks and months, it is assisted by a Global Navigation Satellite Systems (GNSS) timing receiver, which is operated in snapshot mode and activated only for very short time durations of a few seconds on a daily or weekly basis, based on the mission needs. This innovative GNSS assistance replaces the ground-based synchronisation and offers an autonomous operations of the onboard timing system. In this way, a highly accurate yet very low Size, Weight and Power (SWaP) system can be realized, which can be used on very small satellites down to nano-satellites. The onboard architecture of this system is presented and characterised in terms of its achievable Allan deviations as well as its SWaP. It is found that the power consumption can be as low as 2.2 W, based on a daily GNSS synchronisation schedule, significantly reducing the power requirements over current state-of-the-art solutions. An even less complex variant of this innovative system can be employed for any deep space or planetary mission, replacing the onboard GNSS timing receiver by a traditional time synchronisation mechanism using a terrestrial ground station, or, in the further future, using a pulsar.

This paper proposes a novel concept of using highly efficient Snapshot Global Navigation Satellite Systems (GNSS) receivers to provide precise position fixes of single or multiple satellites in Low-Earth Orbit (LEO) to improve upper atmospheric modeling and thus contribute to superior space situational awareness (SSA). While tracking of LEO satellites and the use of onboard GNSS receivers for drag measurements and upper atmosphere modeling are well-established techniques, the expected advent of snapshot GNSS receivers for spaceborne scientific applications will allow massive improvements on the GNSS sensor's Size, Weight, Power and Cost (SWaP-C). With chip-size dimensions of 4x4 mm², a mass of less than 5 gr, an average power level below 0.1 mW, snapshot receiver technology is expected to provide position fixes in space with an accuracy of ∼19 m (3D r.m.s.), which will surpass the accuracy of Two-Line Elements (TLE) provided by the US Joint Space Operations Center (JSpOC) by at least two orders of magnitude. Equally important to their SWaP-C benefits, Snapshot GNSS receivers will allow mission and spacecraft designers to trade onboard-processing requirements versus payload downlink requirements, leading to either minimum onboard processing or a minimum amount of downlinked data. In this research, we establish the concept and architectural overview of using snapshot GNSS receivers for SSA, including the role of using them in a Distributed Space System (DSS), and detail their characterization and performance in terms of the required GNSS hardware and the impact of these payload on the power budget, the link budget and the OnBoard Data Handling (OBDH) budget of a satellite. It will be shown that these receivers lend themselves especially to their use on femto-, pico- and nano-satellites, although integrated snapshot modules may be flown as auxiliary payloads on micro- or mini-satellites as well. While this work focuses on the implications of the use of snapshot GNSS receivers on spacecraft design for the use of upper atmosphere modeling and SSA, their use may open up other science applications which avoid the need for expensive high-grade GNSS receivers.

Line-of-Sight (LoS) navigation is an optical navigation technique that exploits the direction to visible celestial bodies, obtained from an onboard imaging system, to estimate the position and velocity of a spacecraft. The directions are fed to an estimation filter, where they are matched with the actual position of the observed bodies, retrieved from onboard stored ephemerides. As LoS navigation represents a really promising option for the next-generation deep-space spacecraft, the objective of this work is to provide new insights into the performance. First, the information matrix is analyzed to show the influence of the geometry between the spacecraft and the observed planet(s). Then, a Monte Carlo approach is used to investigate the influence of measurement error (ranging from 0.1 to 100 arcsec), and tracking frequency (ranging from four observations per day to one observation every two days). The effect on navigation performance is quantified by two indicators. The first is the 3D position and velocity Root-Mean-Square-Errors, computed once the estimation is considered to be steady-state. The second is the convergence time, which quantifies the required time for the estimation to reach the steady-state behaviour. The simulation is based on a set of four planets, which do not follow the common heliocentric dynamics but rotate around the Sun with the same (distance-independent) angular velocity of the spacecraft. This approach allows the separation of scenario-dependent behaviours from navigation intrinsic properties, as the same relative geometry between observer and observed objects is maintained during the whole simulation. The results provide a useful guide for the next-generation autonomous navigation system, for both the definition of hardware requirements and the design of an appropriate navigation strategy. Considerations are then applied to Near-Earth Asteroid fly-by mission scenarios for the definition of the navigation strategy and hardware requirements. It is shown the importance of relative angles between the spacecraft and the planets. In the single-planet observation scenario, when the angle between the position vectors of the spacecraft and planet approaches a null value, the estimation error decreases. In the double-planets observation scenario, when the separation angle between the two LoS directions gets close to 90°, the estimation error decreases. The main influence on the performance is driven by the measurement error, which with current technologies is shown to be able to provide a position error in the order of a few hundred kilometers, while with a lower measurement error (0.1 arcsec) it would be possible to have a position error below 100 km. Finally, it is demonstrated that tracking frequency plays a secondary role in the performance, and only influences tangibly the convergence time.

Contribution: This article presents a comprehensive overview of characteristics of educational designs of collaborative engineering design activities found in literature and how these characteristics mediate students' collaboration. Background: Engineers have to solve complex problems that require collaboration. In education, various collaborative engineering design activities have been implemented to prepare students for these professional practices. According to cultural historical activity theory (CHAT), educational activities can be described in terms of interrelated elements, i.e., subject, object, tools, rules, division of labor, and community, that influence learning outcomes. A key issue is how these elements mediate students' collaborative efforts and how they contribute to learning. Research Questions: 1) How is collaborative learning implemented in engineering design education? 2) How do the elements of CHAT and their interrelations mediate collaborative learning? and 3) What is the evidence that the implementation of collaborative learning contributed to the achievement of desired learning outcomes? Methodology: A systematic literature review following preferred reporting items for systematic review and meta-analyses protocols guidelines was conducted, including 111 articles published between 2011 and 2021. CHAT was used as analytical framework. Findings: Collaborative learning was implemented in engineering design activities to develop technical as well as nontechnical skills. For the CHAT elements, it was found that establishing a common object, rules for collaboration, and division of labor are essential for effective collaboration and can be enhanced through digital technologies (tools) and support from a community, for example, educators. Finally, results showed that there is evidence that described implementations contribute to learning. However, this evidence needs to be interpreted with care, due to methodological issues in some included articles.

This paper describes the work carried out to extend the NOEL-V platform to include data-level parallelism (DLP) by implementing an integer subset of the RISC-V Vector Extension. The performance and resource utilization efficiency of the resulting vector processor for different levels of DLP (i.e., number of lanes) have been compared to the baseline scalar processor on a Xilinx Kintex Ultrascale field-programmable gate array, employing typical kernels for compute-intensive applications. The role of the memory subsystem has also been investigated, comparing the results obtained with a low-latency and a high-latency main memory. The results show that the speed-up due to the use of the vector pipeline increases with the number of lanes in the vector processor, achieving up to 23.0× the performance of the scalar processor with only 4.3× the resources of the baseline scalar processor. Using an implementation with 32 lanes increases performance even for problem sizes larger than the number of lanes, achieving up to more than 11.7× the performance of the scalar processor with just 1.9× its resource utilization for 128 × 128 matrix multiplications. This work proves that implementations of the selected subset are easily scalable and fit for small-processor implementations in highly constrained space embedded systems.