Artificial intelligence (AI) is adopted at a fast pace and is being applied widely in many domains. The aim of the REXASI-PRO project is to design a novel framework in which safety, security, ethics, and explainability are entangled to develop a Trustworthy Artificial Swarm Intelligence solution. The framework will make a trustworthy collaboration among a swarm formed by autonomous wheelchairs and flying robots to allow a seamless door-to-door experience for people with reduced mobility. Within REXASI-PRO-Hop-On, the widening partner (SIE) will address the following challenges, which were already identified in the REXASI-PRO proposal: - Foundational challenge 3: How to develop green AI systems without affecting the system’s performance? - Domain-specific challenge 4: How to preserve machine learning accuracy while reducing the power consumption of embedded computers? These two challenges are currently being addressed in the REXASI-PRO project by employing optimization and efficiency computational topology methods for energy saving, notably by limiting the amount of training data while maintaining the highest standards of completeness and accuracy. The REXASI-PRO-Hop-On proposal will address a different dimension of a green AI solution: algorithms for AI model compression, with the goal of optimizing the energy efficiency during AI model inference on the wheelchair;

The recently started FLUTE project aims to research and develop a platform for secure, privacy-preserving cross-border federated learning, and includes a pilot study on the diagnosis of prostate cancer. While FLUTE's algorithms will scale well to larger number of patients, it does not include work towards handling very large models, and its pilot study has a limited size of about 3500 patients. In the ALTO-FLUTE project, Siemens Romania (SR) will bring (a) a large dataset with 500 million medical images among which 100.000 prostate MRI images and 12.000 prostate MRI images with diagnosed cancer, and (b) significant expertise in training large-scale foundational models which can exploit not only the more expensive fully labelled data but also the more abundant unlabelled data or incomplete patient records. Supported by the existing partners, SR will be integrated in the FLUTE project, will learn new privacy-enhancing technology, will research and develop new secure, privacy-preserving algorithms, and will add extensive new contributions to the pilot study. This effort will significantly enhance FLUTE's impact, and will stimulate the exchange of scientific knowledge and ideas between Romanian researchers and participants of excellent research projects.

The REXASI-PRO project aims to release a novel engineering framework. The REXASI-PRO project aims to release a novel engineering framework to develop greener and Trustworthy Artificial Intelligence solutions. In the methodology, safety, security, and explainability are entangled. In addition, throughout the entire lifecycle of the framework, ethics aspects will be continuously monitored. To this end, the REXASI-PRO project introduces several novelties. The project will develop in parallel the design of novel trustworthy-by-construction solutions for social navigations and a methodology to certify the robustness of AI-based autonomous vehicles for people with reduced mobility. The trustworthy-by-construction social navigation algorithms will exploit mathematical models of social robots. The robots will be trained by using both implicit and explicit communication. REXASI-PRO methodology augments existing system-level and item-level engineering frameworks by leveraging novel eXplainability methods to improve the entire system's robustness. REXASIPRO will release additional verification and validation approaches for safety and security with the AI in the loop. Among the other developments, a novel learning paradigm embeds safety requirements in Deep Neural Network for planning algorithms, runtime monitoring based on conformal prediction regions, trustable sensing, and secure communication. The methodology will be used to certify the robustness of both autonomous wheelchairs and flying robots. The flying robots will be equipped with unbiased machine learning solutions for people detection that will be reliable also in an emergency. Thus, REXASI-PRO will make the AI solutions greener. To this end, both an AI-based orchestrator to augment the intelligence of the robots and topological methods will be developed. The REXASI-PRO framework will be demonstrated by enabling the collaboration among autonomous wheelchairs and flying robots to help people with reduced mobility.