Institut de Recherche Technologique SystemX

Abstract The integration of artificial intelligence (AI) into safety-critical systems, where human operators remain central to decision-making, introduces various challenges that existing AI frameworks struggle to address comprehensively. Key concerns involve designing a socio-technical system that balances AI transparency, trust, and explainability with the imperative for robust and reliable decision-making. Presently, while numerous sector-specific solutions exist, a holistic framework that effectively integrates human expertise with AI capabilities remains absent, leaving critical gaps in system design, deployment, and oversight. This chapter proposes a multidisciplinary conceptual framework to enhance human-AI collaboration in critical infrastructures such as power grids, railways, and air traffic management. The different design steps were guided by the requirements of these industrial domains. The framework combines key design principles that support human cognition, leveraging insights from decision theory, mathematics, and specialized engineering domains to optimize AI-assisted decision-making. Furthermore, it embeds trustworthiness and risk assessment methodologies, using tools such as the Assessment List for Trustworthy Artificial Intelligence (ALTAI) tool to ensure compliance with ethical and regulatory requirements.

After a brief introduction of the Knowledge Graph (KG) technology, a subfield of symbolic AI used to represent and manage semantic information, the article is devoted to quality assessment, emphasizing the importance of developing trustworthy AI in such knowledge-based systems, particularly in safety-critical applications. In this context, we remind several metrics or methods that can be applied to KGs, along with examples of their implementation in the context of digital NOTAMs illustrated by the HLIF2024 Hackathon.

At the core of Intelligent Transportation Systems (ITS) lies a wide range of applications leveraging communication and information technologies to support safer, more efficient, and more sustainable mobility. However, due to their inherent complexity, ITS architectural design remains challenging, particularly for designers unfamiliar with the domain. A promising approach is to build upon existing ITS reference architectures, which are validated frameworks encapsulating essential architectural knowledge. Yet these reference architectures remain underutilized, in part because the architectural elements and reusable components they provide are insufficiently known and disseminated. To address this gap, we propose a Retrieval-Augmented Generation (RAG) system that enables users to explore such architectures through a question–answer interface. The system combines the generative capabilities of large language models (LLMs) with structured knowledge from established ITS reference architectures, including ARC-IT and FRAME, and has been evaluated from both technical and user-oriented perspectives.

ABSTRACT: Current transitions, such as digital and ecological ones, bring new challenges for organizations, characterized as unknowns. Addressing them requires new management paradigms for which design-based methods show promise. Yet their organizational implementation remains limited, what this paper investigates. Based on a two-year collaboration with a French healthcare company, the study involved developing, delivering and evaluating a four-day training program. Based on interviews and evaluations from 65 participants, results indicate high satisfaction, significant habit disruption and intent to adopt design-based tools. Due to the development of a common language in the organization and the emphasis on learnings’ co-creation, this training had a transformative power. Thus, highlighting its practical value and opening pathways for exploring its long-term impact on organizational practices.

We propose a new methodology for parametric domain decomposition using iterative principal component analysis. Starting with iterative principle component analysis, the high dimension manifold is reduced to the lower dimension manifold. Moreover, two approaches are developed to reconstruct the inverse projector to project from the lower data component to the original one. Afterward, we provide a detailed strategy to decompose the parametric domain based on the low dimension manifold. Finally, numerical examples of harmonic transport problem are given to illustrate the efficiency and effectiveness of the proposed method comparing to the classical meta-models such as neural networks.

An Operational Design Domain (ODD) describes the specific operating conditions in which an autonomous vehicle is designed to operate safely. Although ODD is considered as a central point of knowledge when developing an autonomous system, stakeholders still lack standardized languages and tools for defining an ODD, particularly for non automotive specific domain. This paper proposes a Domain Specific Language (DSL) to define an ODD. The proposed DSL is based on the Ontology Language for the Dependability of Automated Systems (OLDAS) and is independent from the automotive domain. It is implemented in a Model Based System Engineering (MBSE) tool and was successfully used and evaluated on two different use cases. The DSL has been as well used to generate highly concrete scenario from ODD.