ASTIR has completed its first major phase of research and technology development, moving the project from requirements and architectural planning towards implemented prototype components.
During this phase, the consortium developed initial methods and tools across three closely connected areas.
The trustworthy AI activities investigated knowledge-grounded generation, explainability and methods for evaluating generative AI systems. The requirements, design and implementation activities produced early AI-supported workflows for analysing engineering documentation, generating structured requirements, supporting architecture work and assisting software implementation. The dependability activities addressed security review, test generation and runtime monitoring for robotics software.
The project has also established an initial deployment infrastructure through which specialised agents and engineering tools can be integrated. The architecture supports interaction between AI agents, external tools and existing development environments while retaining opportunities for human review and intervention.
The next phase will concentrate on integration and industrial evaluation. Prototype technologies will be connected through the ASTIR workbench and exercised in the manufacturing, mobile service robotics and drone use cases.
The purpose of these evaluations is not only to measure technical performance. They will also examine traceability, robustness, usability and the extent to which engineers can understand, validate and confidently use AI-generated results.
