An integrated approach to smart satellite manufacturing: bridging humans, robots and data

Aerospace manufacturing occupies a distinctive position within industrial production. Space systems are high-value, low-volume artefacts, and their realisation depends on Manufacturing, Assembly, Integration and Testing (MAIT) activities that remain highly procedural and human-centred. In this context, operator expertise is still essential to ensure quality, traceability and process control. The low production volumes, frequent configuration changes and high cost of errors that characterise this domain limit the direct transfer of large-scale automation paradigms from high-volume sectors, shifting the focus from replacing human work to augmenting it. This thesis investigates how Industry 4.0 technologies can be systematically applied to the MAIT of small satellite platforms, with particular attention to CubeSats, along a coherent path from real-time operator support, to knowledge formalisation, to process optimisation. The work is organised around three objectives. First, an XR-assisted Cyber-Physical System framework, driven by a Domain-Specific Language and marker-less object recognition, is designed and evaluated on a CubeSat structural subsystem through a controlled user study with 40 participants, showing improvements in task completion time, error reduction and perceived usability compared to paper-based procedures. Second, the thesis explores the use of an on-premise LLM pipeline to extract procedural knowledge from heterogeneous engineering documents into DSL specifications, and discusses the role of the DSL as a common abstraction layer for integrating XR guidance and collaborative-robot execution. Third, DTLogExtSim is introduced as a containerised, domain-agnostic framework combining automated process discovery, discrete-event simulation and what-if analysis over execution logs to support data-driven decision-making.