AIthena

AI-Based CCAM: Trustworthy, Explainable, and Accountable

Research on Connected and Cooperative Automated Mobility (CCAM) is blooming thanks to Artificial Intelligence (AI), which has made an outstanding breakthrough with the emergence of sophisticated Deep Learning methods, along with the development of automated driving technologies. CCAM solutions have benefited from the applicability of AI-based perception, situational awareness, and decision-making components.

To gain social acceptance, trustworthy AI is the next mandatory step of technology development to deliver the promise of fully beneficial AI. Until recent years, only accuracy has been used as a Key Performance Indicator of AI, as a gold standard to measure AI quality. However, trustworthy AI requires exploring trade-offs among other equally important properties: robustness, privacy, explainability, accountability, and ethics. Explainable AI (XAI) is growing quickly as an area of keen interest for users (citizens to trust the systems they use, legal entities for liability and accountability, and researchers to understand limitations and improve models) of AI systems who are demanding to be provided explanations for AI functioning and expected behaviour. In this context AIthena will tackle methodological and development challenges for the creation and integration of XAI-based models and systems into CCAM applications. AIthena will provide a human-centered methodology aiming towards the evolution of the three main AI pillars: Data, AI-Models and Testing.

ika's main contribution within AIthena is the development of explainable, trustworthy and robust perception and decision-making techniques. Thus, fusion models will be developed that reduce conflicting perception on the one hand and that combine all relevant information for decision-making modules on the other hand. Finally, the resulting modules will be evaluated, tested and integrated, first in simulation and afterwards in demonstrators for both perception- and decision-making.