To achieve this, we combine data-driven edge AI (TinyML) with deterministic expert systems. This allows autonomous decisions based on local knowledge to be made efficiently, safely and traceably.
Focus Areas
Building on previous research, IART is developing a novel system for on-board autonomy. Key areas include:
- decoupling and optimising neural perception and symbolic reasoning
- a "Symbolizer" that acts as an intelligent interface, translating high-frequency data from TinyML models into stable, semantically meaningful facts
- matching system faults against autonomy requirements and the original mission objectives (goal-based decision-making)
- verification and validation (V&V) of the hybrid architecture for inherently explainable AI through rule tracing
- building a representative hardware-in-the-loop (HiL) environment with current space-grade microcontrollers
- sustainable development of robust algorithms for the New Space market
