2025 | Inria Quadrant Program

MIND-SCAN : Intelligent Monitoring for Neuro-Degenerative diseases through connected insoles and self-supervised AI

Benjamin Allaert - Ecole Nationale Supérieure Mines - Télécom Lille Douai

Site PIQ : Université de Lille
Durée et financement : 6 mois - 71.6 k€
Catégorie ANR : biology health

            MIND-SCAN aims to explore the feasibility of inferring pelvic and leg motricity solely from plantar activation data, leveraging self-supervised AI to model human movement while integrating biomechanical constraints. By accurately estimating joint positions, this research could revolutionize non-intrusive monitoring of elderly and disabled individuals.

Neural networks for homomorphic encryption

Jean-Christophe Mourrat - CNRS

Site PIQ : Université ComUE Lyon Saint-Étienne
Durée et financement : 48 months - 288,7 k€
Catégorie ANR : digital sciences and mathematics

            Deep-learning techniques are rapidly advancing but data confidentiality remains a major challenge in sensitive domains like healthcare. Recent cryptographic advances enable computations on encrypted data without decryption, sparking interest in integrating these methods with deep learning. However, conventional neural networks are ill-suited to cryptographic constraints and the goal of this proposal is to make progress on this point.

DTOO : Digital Twins of Organoids

Anaïs Baudot - Aix-Marseille Université

Site PIQ : Aix-Marseille Université
Durée et financement : 48 months - 1 063,8 k€
Catégorie ANR : biology health

            Research on organoids is progressing rapidly but faces challenges in characterization and reproducibility. The DTOO project aims to develop digital twins to unlock the full potential of organoids. To this end, new hybrid AI algorithms will be designed, combining symbolic and numeric approaches to integrate multi-omics multi-scale data.

OSS4EAI : Open-source software for Embodied AI

Stéphane Caron - Inria

Site PIQ : Paris Sciences et Lettres
Durée et financement : 24 months - 866,51 k€
Catégorie ANR : digital sciences and mathematics

            OSS4EAI explores the design of end-to-end differentiable control architectures. It revisits core robotics algorithms, notably for physics simulation, and extends them to automate sim-to-real transfer to real robots with sparse data. The approach will be implemented as maintained open-source software to accelerate learning of new robot behaviors and offer new ways to share data and models.

MIND-SCAN : Intelligent Monitoring for Neuro-Degenerative diseases through connected insoles and self-supervised AI

Neural networks for homomorphic encryption

DTOO : Digital Twins of Organoids

OSS4EAI : Open-source software for Embodied AI

Subscribe to our newsletter