Co-pilots for Arrowhead-based Cyber-Physical System of Systems Engineering

One benefit of Large Language Model (LLM) based applications (e.g. chat assistants or co-pilots) is that they can bring humans closer to the loop in various IT and OT solutions. Co-pilots can achieve many things at once, i.e. provide a context-aware natural language interface to knowledge bases, reach various systems (via APIs), or even help solving multi-step problems with their planning and reasoning abilities. However, making production-grade chat assistants is a topical challenge, as fast-evolving LLMs expose new types of application design and security issues that need tackling. These especially rise to power when we try to apply these solutions to industrial automation use cases, as they need additional explainability and reliability engineered into the architecture. This paper describes the envisioned use cases and the findings of proof of concept copilots for the Cyber-Physical System of Systems (CPSoS) engineering domain. The paper suggests three types of copilots to support the stages throughout the CPSoS engineering lifecycle and shows Proof-of-Concept scenarios for the Eclipse Arrowhead engineering process.