This proposal has been submitted in September 2024 under the Call: HORIZON-CL5-2024-D6-01-04;Type of Action: HORIZON-RIA (HORIZON Research and Innovation Actions).

The AI4TrustDrive proposal is led by Frederick Research Center (FRC).

The primary goal of the AI4TrustDrive project is to apply a novel design methodology to develop a trustworthy, hybrid intelligent compliant automated driving system with Collective Perception Capabilities which fulfils the crucial objective of “calibrating trust” i.e. matching system capabilities with user expectations to promote safe, reliable, efficient and environmentally friendly driving. The project will build on social sciences expertise within the consortium to reveal the factors which mostly affect trust and, henceforth, both control over and adoption of the technology. AI4TrustDrive integrates eXplainable AI (XAI) methodologies, unique to this project the notion of context-specific explanations (i.e. tailored to the situation in which they are being communicated), to ensure calibration of trust as the users have accurate mental models of the system operating capabilities. Trust assessment and calibration will result in user-defined feedback that will be used as a baseline to regulate AI-based “trust-aware” controllers that apply a shared control approach switching between different operating modes of human machine collaboration ensuring hybrid intelligence i.e. smooth integration of human cognitive abilities with AI systems in CCAM systems. Given existing challenges of the collective perception service, the project will also contribute to proposing beyond the state-of-the-art methodologies with which to enhance system reliability and robustness to failures and account for the resulting uncertainties through robust control design and trust supported orchestration of human machine collaboration. The project integrates a novel meta-controller which considers relevant ethical socio-legal values and monitors their compliance. It does so by breaking down relevant policies into normative interpretations and monitors the system’s adherence to them.