The development of the AI4CCAM interoperable digital framework is ongoing and its architecture has been specified during RP1. As part of this digital framework, the AI4CCAM Trustworthy AI methodology has been delivered (D1.1) 12 urban traffic scenarios (functional and logical) have been modelled in MOSAR, and one test protocol is ready for simulation. An initial version of the digital framework has been developed based on the provided results, while its final version will be completed during RP2. Based on these elements R1.1 and R1.3 are completed and R1.2 (second version of the digital framework) is expected at the end of the project.
Trustworthy AI for CCAM through ethical, social and cultural implication is defined and activities are ongoing to improve user acceptance of automated vehicles. The AI4CCAM participatory space (R2.1) is up and running and now expected to be populated by EU citizen contributions. It presents a knowledge base related to AI risks, ethical risks in CCAM and cyber threats for automated vehicles (R2.2). The Ethical and Scientific Advisory Board has been setted up and already met 3 times (R2.3). The production of a roadmap for user acceptance of automated vehicles by VRU is ongoing (R2.4).
Three use cases for exhibiting VRU-AV interactions with potential ethical risks are ongoing. The evaluation criteria related to ethics have been set and are available as part of the AI4CCAM validation handbook (R3.1 D4.1). The CAV-VRU interaction demonstrator using Virtual Reality is well ongoing (R3.2 D4.2). The design of simulation scenarios for the three use cases is ongoing (D4.4). The implementation of the three use cases has started and we are confident that they will be ready to evaluate the multiple research results obtained in the project (R3.3 D4.3).
Research results have been reached with the preparation of multiple scientific articles, that are either published according to open-science practices (at least 6), accepted or submitted for publication (R4.2). Trustworthy-by-design models are being trained or validated with GANs and Metamorphic Testing. These models are designed for pedestrian movement anticipation, scene understanding, and visual gaze estimation (R4.1). Regarding education already two PhD students have been educated on Trustworthy AI for CCAM (R4.3). At mid-term of the project, objective SO4 is well on track and will be reached at the end of the project.
Regarding outreach, many communication and dissemination activities took place during RP1. Apart from the delayed Stakeholders Forum (R5.1°, which was organised in May 2024, instead of Fall 2023, all the planned C&D and exploitation actions have been performed, including a market analysis and first exploitation roadmap (R5.2) and innovation registry (R5.3).
