Launcher manufacturing is undergoing a rapid transformation, driven by the need for increased production robustness, cost reduction, and enhanced sustainability. In particular, the adoption of automation and digitalisation in production processes has become essential to remain competitive in today’s space industry. This thesis contributes to this transition by developing a data-driven anomaly detection system for robotic drilling and riveting, a critical step in the assembly of launcher structures. Drawing inspiration from the field of audio classification, in which transformer-based models have recently achieved state-of-the-art results, this research proposes a novel adaptation of the spectrogram transformer architecture to the domain of industrial time-series sensor data. The resulting deep learning-based approach enables automated detection of anomalous behaviour through sensor signals, offering a scalable solution to process monitoring.

The near-Earth space environment is getting more and more crowded, and this growth in satellite numbers is not expected to slow down. Forecasts report an average of over 3700 satellites to be launched annually. Due to this ever-increasing number of objects in Low-Earth Orbit (LEO), there is a need to improve the accuracy of the positional estimates of spacecraft to better predict close approaches and avoid collisions between operational spacecraft and space debris alike. However, current technologies do not allow global Space Situational Awareness (SSA). This report concerns the design of a small, lightweight, power-efficient and low-cost standardized payload for space traffic management based on a snapshot Global Navigation Satellite System (GNSS) receiver. This standardized payload could then be easily integrated in virtually any LEO spacecraft and provide accurate position fixes for it. This gathered position data would then also enable the improvement of upper atmospheric density models, increasing the precision of LEO object orbit propagation and supporting SSA. From this the team developed the following mission statement: "Design a standardized payload for space traffic management, based on a snapshot GNSS receiver, and an associated technology demonstration mission and satellite to validate this payload." At the core of the standardized payload is a snapshot GNSS receiver, capable of acquiring position data through millisecond-scale signal captures of GNSS signals. This mode of operation enables intermittent, ultra-low power consumption, making it suitable even for power-constrained platforms such as CubeSats or PocketQubes.

At the Deutsches Zentrum für Luft- und Raumfahrt (DLR), a test bench is being designed and constructed. The software and electrical hardware solutions needed present an interesting challenge, particularly in the area of sensor synchronization. Designs for the software and electrical hardware systems are discussed, and then the executed construction of these designs is examined. An additional problem is the issue of verifying that the test setup properly adheres to requirements. A literature gap in space test setup verification is identified, regarding the selection of a rigorous verification approach and the development of verification test plans. In order to address this gap, a method of determining what verification approach should be used, and a systematized format for developing verification test plans is created. In order to determine the efficacy of these methods, the desired test setup is used as a model example. A verification approach for the requirements is made, and then test plans for the selected requirements are written using the developed methodology. These test plans are subsequently executed, and the efficacy of the test plans and their creation methodology are evaluated. From the evaluation criteria for these test plans, it can be seen that the standardized test plan methodology does a good job of ensuring that that requirements and metrologic traceability is maintained, and that there is room for improvement in the isolation of influencing factors.

The Exploration Company's Nyx spacecraft is meant to be the first privately funded European cargo capsule. It will service the growing demand for up- and down-mass to Low Earth Orbit and beyond, supporting both existing systems such as the International Space Station, and future systems developed by companies such as Vast and Axiom. Because Nyx is a complex and strategically significant vehicle, special attention is being paid to the Systems Engineering approaches that are applied to its development. This work presents a novel approach and software platform for organizing, tracking, and analyzing the verification of complex systems. This platform - named "Athena" - enables rapid, incremental verification of these systems, thereby contributing to complete and consistent results. The Exploration Company has begun to apply this platform to Nyx's NASA safety certification, with initially positive results.

While Artificial Neural Networks (ANN) representational capabilities have long been considered to benefit guidance and control tasks, a knowledge gap exists in the literature for their application to augment real-time navigation algorithms. Deep learning has excelled in various tasks, but its computational complexity limits its use in onboard spacecraft navigation. The study focuses on Single Layer Feedforward Networks, simpler ANN structures, to design shallow Physics-Informed Neural Networks (PINNs). The research aims to merge physics-based models with shallow neural networks to improve computational efficiency and enhance autonomy of conventional navigation algorithms. The proposed methodology involves designing a novel architecture combining an Extended Kalman Filter (EKF) with shallow PINNs, while also developing a novel incremental learning technique. Numerical simulations are performed for a relative navigation LEO scenario and results are presented. The PINN-EKF architecture emerges as a robust system, capable of taking advantage of physics-based knowledge, as well as of automatically predicting the dynamics from observations. In particular, it provides substantial computational advantage over standard nonlinear sequential estimators at a limited accuracy and autonomy performance degradation.

Mitigation strategies which eliminate existing space debris, such as with Active Space Debris Removal (ASDR) missions, are now more important than ever regarding the ever-growing space debris population problem. One of the considered ASDR approaches uses a net as a capturing strategy. The benefits of such strategy are to allow for a large capturing distance, high compatibility for different space debris sizes and reduced accuracy requirements.

A key requirement of any ASDR missions is that during capture, no new space debris is to be generated during the process. However, when simulating net capturing in the literature, the potential to break of vulnerable structures, like antennas or solar panels is often neglected. Such elements may show an enhanced risk of failure, especially if these are already damaged, potentially contributing to even more space debris.

A discrete Multi-Spring-Damper net model was used to simulate the 20 m/s-frontal impact of a 30 m x 30 m net onto an ESA Envisat mock-up. The Envisat was modelled as a two rigid-body system with a Single-Degree-of-Freedom hinge connection. A sequential modelling strategy was implemented, which de-coupled all the necessary dynamic and structural models. More than two large sub-structures (the Ka-band antenna dish and solar array) were found to have a high likelihood of breaking, leading to the recommendation of several design mitigation strategies using two types of sensitivity analysis. With secondary space debris being generated, net capturing is found to be riskier than originally assumed throughout the literature.

The exponentially increasing number of Earth-orbiting satellites, as well as the pronounced importance for space military operations, impose new challenges on a functional Space Domain Awareness (SDA). A key factor for an effective SDA is the successful identification of an Earthorbiting satellite, or object in general. Current methods are usually limited to either time-resolved photometry, spectrometry, or polarimetry. In this thesis, we present the Spectropolarimeter for Satellite Identification (SSI), whose design combines these detection methods with the aim to provide a unique ‘fingerprint’ of the satellite, thereby expanding the range of satellite identification methods. The SSI combines a dual-channel spectrometer with off-the-shelf polarization optics. Using a combination of an achromatic quarter-wave retarder, a highly chromatic multiple order retarder, a polarizing beam splitter, and two linear polarizers, all operating in the visible spectrum, a sinusoidal modulation of the linear polarization information is imprinted onto the measured spectrum. The amplitude of the signal scales with the degree of linear polarization and the phase with the angle of linear polarization. Furthermore, by using a dual-channel setup, the intensity spectrum can be obtained for the full spectrograph resolution of 9 nm. We introduce and verify the retrieval algorithm to attain the linear polarization information with an approximate resolution of 135 nm due to the modulation approach. Differential transmission between both channels is corrected by a proposed iterative transmission correction which is verified to have an effect of ≤ 1% on the accuracy of the retrieved linear polarization information. This thesis concentrates on geostationary satellites as observation targets, making use of the stable observation geometry to simplify calibration and analysis. We reason that the phase angle has a strong influence on the retrieved spectropolarimetric data and received photon flux. A thorough photon budget estimation is conducted with the conclusion that we are able to observe faint geostationary satellites with an apparent visual magnitude of 10, especially if the observed phase angle is kept minimal. An analysis of the most probable errors is conducted. We identify the convolution of the instrument response function with the modulated signal and the induced polarization by the telescope and instrument as the most severe ones. Laboratory wavelength and polarimetric calibration are conducted. Using a double glass plate setup to induce low levels of degree of linear polarization we performed laboratory validation measurements. We determined the overall laboratory polarimetric accuracy to be ≤ 1.2%. Additionally, we are able to distinguish signals ≥ 1.3 %. Preliminary observations of three bright stars are conducted. We are able to retrieve instrumentinduced polarization between 1% and 2.5% which is in the expected range. We conclude that there has been a misalignment between the telescope and instrument which causes a large discrepancy between expected and received photons and propose future measures to resolve the issue. The proof of concept to use linear channelled spectropolarimetry for satellite identification with the SSI will open up the way for further technical refinement of the instrument, an extension to low Earth orbit satellites, and the possible addition of phase angle resolved spectropolarimetric measurements to further improve the concept of a satellite‘s spectropolarimetric ‘fingerprint’. Our work is part of a collaboration between Leiden observatory and Delft University of Technology commissioned by the Royal Netherlands Air Force. Access to the telescope and support with the alignment activities were kindly provided by TNO.

The usage of terrestrial processors in space applications is not straightforward, as processors in space face unique challenges due to the effects of the space environment, like ionizing radiation causing Single Event Effects (SEEs). In the nineties, the European Space Agency chose the Scalable Processor ARChitecture (SPARC) Instruction Set Architectures (ISA) for its processors, as it was the only solution available at that time providing both openness and available software support in terrestrial applications. Currently, a large part of the worldwide space community is using SPARC-based radiation-hardened (rad-hard) or radiation-tolerant (rad-tol) LEON processors in ongoing and planned missions, although SPARC processors virtually disappeared from terrestrial applications. Rad-hard and rad-tol processors for space applications typically lag more than a decade behind their commercial counterparts in terms of performance and the gap is widening every year. This is mainly due to the use of Rad-Hard-By-Design (RHBD) cells and older technology nodes. The larger vulnerability to SEEs of complex microarchitectures is not the only reason why simple microarchitectures with low parallelism are still the vast majority of processors employed in space. As a matter of fact, most of the tasks executed by processors in space data systems are non-compute-intensive workloads. The reason is that they are mainly employed for non-demanding control and housekeeping operations. Therefore, enabling demanding tasks, such as the execution of Artificial Intelligence (AI) algorithms in space embedded systems, requires a large leap in spacegrade processors, especially because space data systems in satellites are typically powerconstrained. Recently, RISC-V, a novel free and open ISA, has risen in popularity in terrestrial applications, drawing the attention of several universities and companies. Given the similarity between SPARC and RISC-V, this dissertation starts by analyzing the advantages of using RISC-V in space applications. The openness of RISC-V already enabled a vast field of research activities for terrestrial applications, with many tools and models at different level of abstraction already available. Therefore, the space industry can spin-in developments from academia and industry, focusing efforts mainly on improvements concerning specific needs in space applications and without wasting efforts on other activities. In order to fully exploit modularity, the need of defining the types of processors required in space application was identified in this dissertation. The modularity of RISC-V was employed to identify several applications in space data systems and RISC-V processor profiles to address them. They were defined in this work by the ISA subset, Instruction-Level Parallelism (ILP), Data-Level Parallelism (DLP), Processor- Level Parallelism (PLP), reference implementation and expected performance. The processors profiles defined range from microcontrollers to general-purpose implementations to high-performance processors for AI. Finally, a roadmap to bring RISC-V IP cores for terrestrial applications to space level was defined, identifying the steps and models required. After the thorough analysis of the state-of-the-art of RISC-V processors was completed, two different sets of activities were identified…

The paradigm of swarm robotics aims to enable several independent robots to collaborate together toward collective goals. The distributed nature of a swarm, whereby each robot acts independently in accordance with its perceived environment, is expected to provide the system with a high degree of flexibility, robustness, and scalability. However, this comes at the cost of increased system complexity. This thesis explores how to automatically design a collective behavior in a way that is transparent and verifiable. The thesis begins by taking a step back and analyzing the design choices that need to be made when designing a swarm of robots. Through an in-depth literature study, focusing on swarms of small drones as a case study, we found how sensor and actuator choices can create constraints for the swarm behavior that can be achieved, and how desired swarm behaviors can create requirements for the hardware design and local-level controllers. Coincidentally, we found a prominent example of this in our own research on relative localization sensors for swarms of tiny drones (performed in addition to the research in this thesis), whereby we developed a communication-based relative localization approach that enabled teams of tiny drones to fly together in tight areas, the advantages being: omni-directional sensing, independence from lighting conditions and/or visual clutter, low mass, and low computational costs. However, this solution also comes with the restriction of ensuring that robots never move parallel to each other, as this will present an unobservable situation. Based on such lessons, the remainder of the thesis aims for a framework that is agnostic with respect to the robot and the swarm's collective task. The framework proposed in this thesis is centered around the following notion: a collective goal can be broken down into a set of locally observable objectives which the robots can sense, referred to as ``desired'' objectives. The robots then take actions in order to reach these desired objectives. When all robots achieve the desired objectives, then the global goal and/or collective behavior emerges. This framework was first developed for the specific case study of pattern formation by cognitively limited robots, which could only sense the relative location of close-by neighbors. It was later generalized, and its use was demonstrated on other collective tasks, namely: aggregation, consensus, and foraging. Through a local model of agent transitions, it was possible to: 1) identify potential obstructions to achieving the collective goal, and 2) optimize the behavior of the robots so as to maximize the likelihood of achieving the desired objectives. The optimization is performed by an evolutionary algorithm that leverages the local model, whereby the fitness function maximizes the probability of being in a desired local state. Using this approach, the policy evaluation only scales with the size of the local state space, and demands much less computation than swarm simulations would. In the final stage of this research, a complete framework was further developed to alleviate the need to manually define the desired objectives as well as the local models required for potential verification and/or optimization. The framework uses a data-driven approach to automatically extract two models: 1) a deep neural network that estimates the global performance of the swarm from the distribution of local sensor data, and 2) a probabilistic state transition model that explicitly models the local state transitions (i.e., transitions in observations from the perspective of a single robot in a swarm) given a policy. The framework can efficiently lead to effective controllers, as demonstrated via multiple case studies. It can also be used in combination with an evolutionary optimization process, leading to higher efficiency, or for heterogeneous online learning. Overall, the methods and insights developed in this thesis propose a new way to approach the development of verifiable and understandable behaviors for swarms of robots, using models in order to perform analysis, verification, and optimization.

Exploring and utilising space to benefit humankind is undeniably a costly venture where the higher the mass of the satellite, the higher the cost to launch. However, the rewards of space-related achievements are indisputable. These achievements range from increasing our understanding of the universe using satellites, such as the Hubble Space Telescope,to improving the predictive ability of weather, climate, and natural disaster phenomena with the help of Earth observation satellites.

Since the launch of the first artificial satellite in October 1957, both satellite computers and their onboard software have changed significantly to integrate more functionalities and making mission operations more reliable. As a result, satellites have become more sophisticated and mission designers have moved functionality from the ground segment to the satellite onboard computers to make them more autonomous. In fact, a larger number of satellite components are adopting the use of embedded computers to improve their performance and increasing subsystems miniaturization. This can be seen by the growth in the number of small satellite missions ranging from 1 kg to 100 kg of mass launched during the past 10 years.

Rendezvous in orbit has recently regained considerable attention, as it is required to enable on-orbit servicing or active debris removal activities. The pressing need for the realization of suchmissions falls within the more general societal attempt to make human activities more sustainable, avoiding wasting valuable resources and ensuring that the environment remains clean after exploitation. Despite the technical heritage of decades of experience, space rendezvous faces, with these new prospects, additional challenges due to the possible noncooperative nature of the target of the rendezvous. A successful and safe approach has to be ensured with limited relative navigation capabilities while reducing the overall mission costs. This quest for cost-effectiveness is indeed required to eventually reach an economically viable large-scale solution able to mitigate the threat posed by the evergrowing population of orbiting space debris.
This dissertation demonstrates that the first part of a rendezvous to a noncooperative object, starting from large separations of several tens of kilometers down to a few hundred meters, can be safely and reliably performed using line-of-sight navigation and solely relying on a single spaceborne camera. More specifically, this research shows that it is possible to use a simple, low-cost, computationally-light and autonomous camerabased embedded navigation system to perform the far-to mid-range approach, thus greatly reducing the necessary onboard equipment and the operational costs. In order to demonstrate this assertion, the dissertation is articulated around three Research Questions:
How to design a reliable and accurate spaceborne real-time angles-only relative
navigation system? How does it behave under real conditions? How can future anglesonly relative navigation systems be improved?

With the rising demands from customers and users and the development of ever advanced technologies, many space missions nowadays require more than one satellite to fulfill their mission objectives. Although replacing single satellite systems (SSSs) by multiple satellite systems (MSSs) offers advantages, such as enhanced spatial and temporal coverage as well as high robustness and multifunctional purposes, it also introduces new challenges. There is no doubt that as the number of satellites in a mission grows, the complexity and operation cost of controlling and coordinating these satellites only by human (or ground based) operators will increase dramatically. In addition, for some deep space missions or complex operational tasks, due to the long signal transmission time between the spacecraft and ground-based antennas or short communication windows, there will not be enough time or resources for operators to sufficiently and efficiently control all of the required onboard functions from mission control centers. Therefore, to enhance the efficiency of operating an MSS, and to reduce the cost of human resources and ground infrastructure, an onboard autonomous system (OAS) for MSS is a promising solution. For specific missions, the use of an OAS may even be a mission enabler. One important function of an OAS is to provide planning and re-planning services based on different mission requirements. The objective of this research is to develop and characterize onboard autonomous mission planning and re-planning approaches for MSSs. Traditional planning approaches have been reviewed and proven to be inappropriate and inefficient for complex planning problems in the harsh space environment when severe system constraints are enforced and a large number of vehicles constitutes the MSS. % Artificial intelligence (AI) approaches, in contrast, are more suitable for complex problems due to their broad adaptability and their ability to cope with large-scale variables. To overcome these deficiencies, engineers and researchers have started to develop OAS with the help of Artificial Intelligence (AI) techniques to allow for more complex space missions. Based on the relevance of this problem, the following research questions (RQs) have been formulated and will be answered in this thesis. % and a review of the state-of-the-art scientific literature \textbf{RQ1: What are the strengths of using AI in space missions? How to use a centralized AI algorithm in a multi-satellite system to decompose mission objectives and perform mission planning for the entire system?} \textbf{RQ2: How to define emergency situations which may occur during mission operations? How to use AI algorithms to handle mission re-planning and re-scheduling problems?} \textbf{RQ3: How to design cooperation and negotiation approaches for an MSS to reach an agreement? How to improve AI algorithms for distributed onboard mission planning problems?} To define potential scenarios, a reference mission is introduced in this thesis, called \textit{Discovering the Sky at the Longest Wavelength (DSL)}. The mission is assumed to comprise one Mother Satellite (MS) and eight Daughter Satellites (DSs) in a lunar orbit. Its scientific objective is to observe the universe in the hitherto-unexplored very low frequency (below 30 MHz) electromagnetic spectrum. The DSs collect scientific data only in those parts of the orbit which is shielded from radio frequencies emitted by the Earth. These DSs can only transmit collected data to the MS when they are outside of this shielded orbit sections, to prevent interferences caused by communication. % The DSs collect scientific data and transmit those to the MS with the constraints what scientific data collection may only occur in the part of the orbit. This part of the orbit is shielded from radio frequencies emitted by the Earth and no other DSs are transmitting data to the MS. This renders mission operations of DSL very complex. The existing body of knowledge on mission planning problems for multi-satellite systems is reviewed. It comprises three categories: classical approaches, heuristic approaches, and advanced techniques (e.g., team negotiation mechanisms, evaluation algorithms). Targeting the complexity of foreseeable DSL planning problems, nine representative optimization algorithms are applied to fourteen test functions. The results indicate that Evolutionary Algorithms (EAs) have a broader adaptability than classical approaches. They are also more efficient than other heuristic approaches. Therefore, EAs family is selected as suitable candidate for the reference MSS. % fourteen test functions are used to test nine representative algorithms to provide a preliminary selection for the reference MSS.=-098 % The fourteen test functions we used contain different types of objective functions and constraints to guarantee the diversity of the preliminary selection. % The goal of this selection is to identify a suitable approach for an MSS to handle different types of optimization problems. Eight constrained and six unconstrained test functions are employed as benchmarks. The operations concept of the DSL mission foresees that the initial mission planning is performed by the MS, while the eight DSs are preliminary executing data collection and transmission tasks. % Considering the scientific design of the DSL mission, the initial mission planning procedures are all performed by the Mother Satellite (MS), while the rest eight Daughter Satellites (DSs) are just participating satellites which response for gathering and transmitting data. During this phase, the MSS implements a centralized architecture and the MS conducts a centralized planning approach. By comparing basic Genetic Algorithm (GA) with several state-of-the-art improved GAs, its weaknesses are revealed. In this thesis, to overcome early and slow convergence problems, the need to develop a new mutation strategy for GA is motivated. % By revealing the weaknesses of the basic Genetic Algorithm (GA), along with a comparison with several other improved GAs, The proposed novel mutation strategy is called Hybrid Dynamic Mutation (HDM), which contains a standard mutation operator and an escape mutation operator. While the standard mutation operator uses a small mutation rate for approaching the global optimum, the escape mutation operator uses a larger mutation rate to allow an escape from local optima. The simulation results indicate that the proposed HDM can improve the basic GA (which turns into the HDMGA) leading to a superior performance on correctness and effectiveness as compared to alternative GAs. Based on these findings, AI related methods are considered a promising category as compared to classical methods due to their flexibility and effectiveness to support the onboard planning for an MSS. In addition, the proposed HDMGA also provides a satisfying result for the considered initial mission planning problems. Internal or external causes, e.g. an actuator failure or the challenging space environment, can lead to a satellite malfunction during mission operations. This thesis considers the two most important behaviors of the DSL mission, observation and communication, and proposes potential emergency scenarios to handle possible system failures on DSs. Two re-planning methods, one called the Cyclically Re-planning Method (CRM), the other one the Near Real-time Re-planning Method (NRRM), are established and compared. The CRM performs re-planning at the beginning of each orbit and only re-plans for one orbit. The NRRM performs re-planning in a near real-time setting when the emergency occurs. Its re-planning covers for the rest of the mission. Three simulation study cases are formulated based on assumed emergency scenarios. % Each case is designed to represent a different level of failures on multiple DSs. The proposed two methods are compared on three aspects: the total number of data observed from all DSs within a certain time frame, the total number of data the MS received from all DSs within a certain time frame, and the average computation time needed for re-planning. The results indicate that: (1) The NRRM allows to observe and transmit more data than the CRM within a specific operational lifetime. (2) The NRRM requires more computational time than the CRM for emergency situations, while it requires less time than the CRM for nominal situations. % For emergency situations, the CRM can therefore provide re-planning sequences faster than the NRRM, while for nominal operations, the NRRM is much faster than the CRM. This research also covers a much more severe scenario, namely that the MS becomes fully non-functional in an emergency situation. This would render the MS unable to provide mission planning and re-planning services for the MSS. Without its main controller on the MS, all DSs now need to cooperate to jointly solve the mission planning problems. Due to the loss of the MS, both distributed and decentralized architectures, which the MSS could then use are introduced. In a distributed architecture, each DS is connected with all other DSs directly or through DS which acts as retranslator. In a decentralized architecture, each DS can only communicate with its neighbors. Considering that the mission allocation problems in different organizational architectures are similar to information games in game theory, a game-theoretical model of the Multi-Satellite Mission Allocation (MSMA) problem is formulated. % Three new negotiation mechanisms are introduced, compared and analyzed in theoretical terms. The Utility-based Regret Play (URP) negotiation mechanism is proposed for an MSMA problem using a distributed architecture. It inherits the ability to evaluate individual utility at each negotiation step from the Utility-based Fictitious Play, and the ability to regret the current choice and for not proposing particular choices in the past negotiation steps from the Regret Matching Play. The Smoke Signal Play (SSP) and Broadcast-based Play (BBP) are developed for a decentralized architecture instead. The SSP is inspired by an old communication method called \textit{Smoke Signal}, where each satellite is considered as a smoke tower, passing information of utility to its succeeding neighbor. The BBP uses broadcasting as the communication method, where each satellite can transmit information to all its neighbors. The simulation results show that the URP can outperform the other three state-of-the-art mechanisms (Action-based Fictitious Play, Utility-based Fictitious Play, and Regret Matching Play) with studied cases. For the decentralized architecture, the results reveal that both SSP and BBP can provide valid solutions for mission allocation problems. The BBP mechanism shows a superior performance on computation time as compared to SSP and a state-of-the-art approach called Market-based Auction. The SSP mechanism, on the other hand, shows the best performance with respect to power consumption. To solve complex optimization problems in distributed mission planning scenarios, an approach, named Hybrid Distributed GA (HDGA) is proposed. This approach contains two modules: the Local Constraint Satisfaction module (LCS) and the Globally Distributed Optimization module (GDO). In the LCS module, the greedy best-first search algorithm is employed as the local search heuristic for helping each DS to find suitable solutions which can satisfy individual constraints. This module is designed to generate multiple solutions to form local populations for the GDO. The GDO module employs the HDMGA as the core optimization algorithm, while the individual populations are formed through the local populations exchange procedure between one participant and all other participants. For a standard planning case, the results indicate that HDGA can reduce the computation time while ensuring a higher success rate compared to the HDMGA. Comparing the HDGA with two other state-of-the-art distributed optimization algorithms, the Distributed Ant Colony Optimization (DACO) and the Coevolutionary Particle Swarm Optimization (CPSO), the statistical results reveal that HDGA is more stable and accurate to handle large-scale planning problems. The HDGA also shows the best performance on computation time among all tested distributed approaches. With the rising demands from customers and users and the development of ever advanced technologies, many space missions nowadays require more than one satellite to fulfill their mission objectives. Although replacing single satellite systems (SSSs) by multiple satellite systems (MSSs) offers advantages, such as enhanced spatial and temporal coverage as well as high robustness and multifunctional purposes, it also introduces new challenges. There is no doubt that as the number of satellites in a mission grows, the complexity and operation cost of controlling and coordinating these satellites only by human (or ground based) operators will increase dramatically. In addition, for some deep space missions or complex operational tasks, due to the long signal transmission time between the spacecraft and ground-based antennas or short communication windows, there will not be enough time or resources for operators to sufficiently and efficiently control all of the required onboard functions from mission control centers. Therefore, to enhance the efficiency of operating an MSS, and to reduce the cost of human resources and ground infrastructure, an onboard autonomous system (OAS) for MSS is a promising solution. For specific missions, the use of an OAS may even be a mission enabler. One important function of an OAS is to provide planning and re-planning services based on different mission requirements. The objective of this research is to develop and characterize onboard autonomous mission planning and re-planning approaches for MSSs. Traditional planning approaches have been reviewed and proven to be inappropriate and inefficient for complex planning problems in the harsh space environment when severe system constraints are enforced and a large number of vehicles constitutes the MSS. % Artificial intelligence (AI) approaches, in contrast, are more suitable for complex problems due to their broad adaptability and their ability to cope with large-scale variables. To overcome these deficiencies, engineers and researchers have started to develop OAS with the help of Artificial Intelligence (AI) techniques to allow for more complex space missions. Based on the relevance of this problem, the following research questions (RQs) have been formulated and will be answered in this thesis. % and a review of the state-of-the-art scientific literature \textbf{RQ1: What are the strengths of using AI in space missions? How to use a centralized AI algorithm in a multi-satellite system to decompose mission objectives and perform mission planning for the entire system?} \textbf{RQ2: How to define emergency situations which may occur during mission operations? How to use AI algorithms to handle mission re-planning and re-scheduling problems?} \textbf{RQ3: How to design cooperation and negotiation approaches for an MSS to reach an agreement? How to improve AI algorithms for distributed onboard mission planning problems?} To define potential scenarios, a reference mission is introduced in this thesis, called \textit{Discovering the Sky at the Longest Wavelength (DSL)}. The mission is assumed to comprise one Mother Satellite (MS) and eight Daughter Satellites (DSs) in a lunar orbit. Its scientific objective is to observe the universe in the hitherto-unexplored very low frequency (below 30 MHz) electromagnetic spectrum. The DSs collect scientific data only in those parts of the orbit which is shielded from radio frequencies emitted by the Earth. These DSs can only transmit collected data to the MS when they are outside of this shielded orbit sections, to prevent interferences caused by communication. % The DSs collect scientific data and transmit those to the MS with the constraints what scientific data collection may only occur in the part of the orbit. This part of the orbit is shielded from radio frequencies emitted by the Earth and no other DSs are transmitting data to the MS. This renders mission operations of DSL very complex. The existing body of knowledge on mission planning problems for multi-satellite systems is reviewed. It comprises three categories: classical approaches, heuristic approaches, and advanced techniques (e.g., team negotiation mechanisms, evaluation algorithms). Targeting the complexity of foreseeable DSL planning problems, nine representative optimization algorithms are applied to fourteen test functions. The results indicate that Evolutionary Algorithms (EAs) have a broader adaptability than classical approaches. They are also more efficient than other heuristic approaches. Therefore, EAs family is selected as suitable candidate for the reference MSS. % fourteen test functions are used to test nine representative algorithms to provide a preliminary selection for the reference MSS.=-098 % The fourteen test functions we used contain different types of objective functions and constraints to guarantee the diversity of the preliminary selection. % The goal of this selection is to identify a suitable approach for an MSS to handle different types of optimization problems. Eight constrained and six unconstrained test functions are employed as benchmarks. The operations concept of the DSL mission foresees that the initial mission planning is performed by the MS, while the eight DSs are preliminary executing data collection and transmission tasks. % Considering the scientific design of the DSL mission, the initial mission planning procedures are all performed by the Mother Satellite (MS), while the rest eight Daughter Satellites (DSs) are just participating satellites which response for gathering and transmitting data. During this phase, the MSS implements a centralized architecture and the MS conducts a centralized planning approach. By comparing basic Genetic Algorithm (GA) with several state-of-the-art improved GAs, its weaknesses are revealed. In this thesis, to overcome early and slow convergence problems, the need to develop a new mutation strategy for GA is motivated. % By revealing the weaknesses of the basic Genetic Algorithm (GA), along with a comparison with several other improved GAs, The proposed novel mutation strategy is called Hybrid Dynamic Mutation (HDM), which contains a standard mutation operator and an escape mutation operator. While the standard mutation operator uses a small mutation rate for approaching the global optimum, the escape mutation operator uses a larger mutation rate to allow an escape from local optima. The simulation results indicate that the proposed HDM can improve the basic GA (which turns into the HDMGA) leading to a superior performance on correctness and effectiveness as compared to alternative GAs. Based on these findings, AI related methods are considered a promising category as compared to classical methods due to their flexibility and effectiveness to support the onboard planning for an MSS. In addition, the proposed HDMGA also provides a satisfying result for the considered initial mission planning problems. Internal or external causes, e.g. an actuator failure or the challenging space environment, can lead to a satellite malfunction during mission operations. This thesis considers the two most important behaviors of the DSL mission, observation and communication, and proposes potential emergency scenarios to handle possible system failures on DSs. Two re-planning methods, one called the Cyclically Re-planning Method (CRM), the other one the Near Real-time Re-planning Method (NRRM), are established and compared. The CRM performs re-planning at the beginning of each orbit and only re-plans for one orbit. The NRRM performs re-planning in a near real-time setting when the emergency occurs. Its re-planning covers for the rest of the mission. Three simulation study cases are formulated based on assumed emergency scenarios. % Each case is designed to represent a different level of failures on multiple DSs. The proposed two methods are compared on three aspects: the total number of data observed from all DSs within a certain time frame, the total number of data the MS received from all DSs within a certain time frame, and the average computation time needed for re-planning. The results indicate that: (1) The NRRM allows to observe and transmit more data than the CRM within a specific operational lifetime. (2) The NRRM requires more computational time than the CRM for emergency situations, while it requires less time than the CRM for nominal situations. % For emergency situations, the CRM can therefore provide re-planning sequences faster than the NRRM, while for nominal operations, the NRRM is much faster than the CRM. This research also covers a much more severe scenario, namely that the MS becomes fully non-functional in an emergency situation. This would render the MS unable to provide mission planning and re-planning services for the MSS. Without its main controller on the MS, all DSs now need to cooperate to jointly solve the mission planning problems. Due to the loss of the MS, both distributed and decentralized architectures, which the MSS could then use are introduced. In a distributed architecture, each DS is connected with all other DSs directly or through DS which acts as retranslator. In a decentralized architecture, each DS can only communicate with its neighbors. Considering that the mission allocation problems in different organizational architectures are similar to information games in game theory, a game-theoretical model of the Multi-Satellite Mission Allocation (MSMA) problem is formulated. % Three new negotiation mechanisms are introduced, compared and analyzed in theoretical terms. The Utility-based Regret Play (URP) negotiation mechanism is proposed for an MSMA problem using a distributed architecture. It inherits the ability to evaluate individual utility at each negotiation step from the Utility-based Fictitious Play, and the ability to regret the current choice and for not proposing particular choices in the past negotiation steps from the Regret Matching Play. The Smoke Signal Play (SSP) and Broadcast-based Play (BBP) are developed for a decentralized architecture instead. The SSP is inspired by an old communication method called \textit{Smoke Signal}, where each satellite is considered as a smoke tower, passing information of utility to its succeeding neighbor. The BBP uses broadcasting as the communication method, where each satellite can transmit information to all its neighbors. The simulation results show that the URP can outperform the other three state-of-the-art mechanisms (Action-based Fictitious Play, Utility-based Fictitious Play, and Regret Matching Play) with studied cases. For the decentralized architecture, the results reveal that both SSP and BBP can provide valid solutions for mission allocation problems. The BBP mechanism shows a superior performance on computation time as compared to SSP and a state-of-the-art approach called Market-based Auction. The SSP mechanism, on the other hand, shows the best performance with respect to power consumption. To solve complex optimization problems in distributed mission planning scenarios, an approach, named Hybrid Distributed GA (HDGA) is proposed. This approach contains two modules: the Local Constraint Satisfaction module (LCS) and the Globally Distributed Optimization module (GDO). In the LCS module, the greedy best-first search algorithm is employed as the local search heuristic for helping each DS to find suitable solutions which can satisfy individual constraints. This module is designed to generate multiple solutions to form local populations for the GDO. The GDO module employs the HDMGA as the core optimization algorithm, while the individual populations are formed through the local populations exchange procedure between one participant and all other participants. For a standard planning case, the results indicate that HDGA can reduce the computation time while ensuring a higher success rate compared to the HDMGA. Comparing the HDGA with two other state-of-the-art distributed optimization algorithms, the Distributed Ant Colony Optimization (DACO) and the Coevolutionary Particle Swarm Optimization (CPSO), the statistical results reveal that HDGA is more stable and accurate to handle large-scale planning problems. The HDGA also shows the best performance on computation time among all tested distributed approaches.

KNO3-Sugar propellants, also known as rocket candy, form a group of simple, cheap and safe solid propellants that are used extensively in student and amateur rocketry communities. One of the most frequently used compositions is KNO3-sorbitol (KNSB) with a typical 65/35 ratio by mass. This composition is used extensively by Delft Aerospace Rocket Engineering (DARE) for small experimental launches including the record-breaking flight of Stratos I in 2009 to 12.3 km altitude. However, KNSB propellant quality has been inconsistent: propellant density is occasionally below 85% in combination with large surface defects. Besides a very high grain rejection rate this has resulted in several explosive failures of new experimental motors in recent years.

The PocketQube is an emerging satellite class, which pushes the miniaturization of space technology beyond the well-established CubeSats, promising rapid design-to-orbit cycles while lowering the cost of accessing space. A showstopper in the success story of nano- and picosatellites are their high mission failure rates. Environmental testing before flight is an effective means to identify design flaws and workmanship errors, and thus improve the chances of mission success. However, inefficiencies in the design process, low-budgets, and stringent schedule requirements often motivate small satellite developers to postpone environmental testing towards the end of the design lifecycle, where recovering for design flaws is inefficient. The project circumstance of nano- and picosatellite missions require, therefore, cost- and time-effective test strategies that facilitate early design evaluation. This research proposes and implements a thermal screening method for PocketQube subsystems to identify temperature hotspots and verify their compliance against operational hardware limits. Key elements of the test method are a thermal IR temperature scan at ambient conditions and an estimation of the worst-case flight temperature using experimentally derived graphs that describe the vacuum heating of thermal hotspots. The study of subsystem layout options complements the screening method by providing solutions to mitigate hotspot overheating. Moreover, the study proposes to lower the required pressure levels for thermal-vacuum testing of PocketQube subsystems. The analysis shows that, due to the small form factor, pressure levels by four orders of magnitude larger than those used in environmental test standards for larger satellites suffice to maintain the resulting temperature errors below 5K on the hot side of the temperature spectrum. Both the screening method and moderation in vacuum requirements contribute to the development of subsystem test methods that match the needs of small satellite developers.

The aerospace industry is recently experiencing growing interest in very small spacecraft like nano- and pico-satellites. However, these very small satellites are still being developed in most cases without a dedicated propulsion system limiting their capabilities. The micro-resistojet has been recognized as a suitable propulsion system for these classes of satellites due to its scalability and performance. Additionally, it can be classified as a "green" propulsion system since it can use naturally any kind of propellant, including "green" propellants. The Low-Pressure Micro-Resistojet (LPM) is a type of micro-resistojet concept that works under very low pressure. This PhD thesis is focussed on the development of this propulsion system concept with the goal to enable very small satellites to perform manoeuvres. This improvement allows, for instance, to increase the spacecraft lifetime by active orbit keeping. Furthermore it can enable orbit change manoeuvres and formation flight.

This thesis reports on the reconstruction of key orbital elements of the Earth’s orbit around the Sun from a miniaturized temperature sensor onboard of theDelfi-C3 Cubesat, a novel approach never explored before to the best of our knowledge. Delfi-C3 is a triple-unit CubeSat with a mass of 2.4 kg, developed by Delft University of Technology, which was launched on April 28th in 2008. Despite of its required lifetime of less than four months, Delfi-C3 is still operational and has been beaconing data for 8.6 years, which makes this CubeSat mission unique in terms of a continuous record of telemetry data. Recent inspections of Delfi-C3 telemetry data over five years from different temperature sensors have revealed surprisingly systematic patterns of periodic nature with an amplitude of 3.1 K and a period of one synodic year. First speculations associated this behavior to be correlated to the orbit of the Earth around the Sun. An analytical model of the temperature fluctuations has been established which associates the amplitude, phase and period of the observed temperature to the eccentricity, the argument of periapsis, and the period of the Earth’s orbit around the Sun, respectively. This analysis represented the first attempt to reconstruct the Earth’s orbit from satellite temperature measurements. To quantitatively estimate the Earth orbit parameters from in-flight telemetry data, a numerical least-squares estimator has been developed and applied to Delfi-C3 temperature data over the period 2008-2015. Preliminary results by using temperature data from the On Board Computer (OBC) Printed Circuit Board already showed that the estimated semi-major axis, eccentricity, and argument of periapsis of the Earth orbit differ from the true values of less than 1%, 3%, and 5%, respectively. An additional filtering of raw data further demonstrated that the achievable accuracy of the estimation can be refined up to better than 0.05%, 2% and 2.5%, respectively. In this thesis, the sensitivity of those results to initial conditions, filtering schemes and estimator settings are addressed. Furthermore, a refined analytical thermal model is created for
Delfi-C3 internal stack, in order to check if a comparable accuracy can be obtained also for other instruments that are less sensitive to external temperature fluctuations, in cases where the internal dissipation requires a better representation of the thermal paths in the internal stack.
In addition to the above analysis, the attitude determination systemon board Delfi-C3 is also reviewed by analyzing the telemetry data fromthe AutonomousWireless Sun Sensor (AWSS) and from the four photodiodes
located on the deployable solar panels. By adopting an analytical model for the Earth albedo, the possibility of retrieving a three-axis attitude determination from these simple sensor is assessed for the case on hand. This
ismade bymonitoring the influence of the albedo termon the determination accuracy for several periods in Delfi-C3 telemetry. It is expected that this analysis will contribute to the understanding of the accuracies of a coarse three-axis attitude determination that makes use of only Sun sensors. Furthermore, the results can be used to determine key factors in the hysteretic process in-orbit, which occur when a PassiveMagnetic Attitude
Stabilization (PMAS) is chosen, by comparing the analytical models for the satellite dynamics with the in-orbit attitude obtained from the deterministic method herewith reviewed. This is expected to contribute to a better understanding and improvement of models used in the design of future CubeSat missions.