Runtime Root-Cause Analysis For Intelligent Robots Via Causal ai Techniques H/F - 95

Le CEA est un acteur majeur de la recherche, au service des citoyens, de l'économie et de l'Etat.

Il apporte des solutions concrètes à leurs besoins dans quatre domaines principaux : transition énergétique, transition numérique, technologies pour la médecine du futur, défense et sécurité sur un socle de recherche fondamentale. Le CEA s'engage depuis plus de 75 ans au service de la souveraineté scientifique, technologique et industrielle de la France et de l'Europe pour un présent et un avenir mieux maîtrisés et plus sûrs.

Implanté au coeur des territoires équipés de très grandes infrastructures de recherche, le CEA dispose d'un large éventail de partenaires académiques et industriels en France, en Europe et à l'international.

Les 20 000 collaboratrices et collaborateurs du CEA partagent trois valeurs fondamentales :

- La conscience des responsabilités
- La coopération
- La curiositéRoot-Cause Analysis (RCA) is a systematic process for identifying the fundamental cause of a problem or failure, rather than merely addressing its symptoms. It aims to understand why something went wrong in order to take appropriate actions and prevent recurrence. RCA is essential for robots that operate outside strictly controlled environments, where they are inevitably confronted with unexpected situations and failures. Symptoms can range widely, including erratic movements, sudden halts, or suboptimal task outcomes. RCA distinguishes these symptoms from the actual causes, which may include hardware or software bugs, inaccurate behavior specifications, or environmental factors. By pinpointing the root cause, robots can select appropriate goals for repair or system adjustments. This informed decision-making enhances resilience and ensures long-term safe autonomy for robots.
Causal inference is a branch of AI research that focuses on understanding and modeling cause-and-effect relationships, unlike many conventional machine learning approaches that primarily seek to identify patterns or correlations within data without establishing causal directions. The primary objective of the internship is to investigate and experiment with the application of causal AI techniques to develop runtime RCA capabilities for intelligent robots. The candidate will survey various approaches from the scientific literature, select a few that appear most suitable for runtime RCA, and conduct experiments to analyze and compare them by utilizing and customizing existing software implementations. The experiments will be conducted in simulated scenarios, with the potential to transition to a physical setup.
The internship covers the following activities:
Conduct a survey of causal AI techniques from the scientific literature (e.g., Bayesian network-based methods, counterfactual reasoning, etc.), with a focus on those applicable to runtime RCA in intelligent robots.

Select a few promising approaches based on the modeling assumptions that characterize the simulated scenarios.

Choose an open-source software framework from among the many existing ones that support the selected approaches (e.g., PyMC, CausalNex, DoWhy, etc.).

Conduct experiments in simulated scenarios to analyze and compare the performance of the selected causal AI approaches in diagnosing and resolving anomalies at runtime, and envision how they could complement or be complemented by other tools and approaches.

Implement a ROS 2 stack that wraps the implemented runtime RCA capabilities.

[Optional] Apply the implemented runtime RCA capabilities to a physical setup, if time permits.

Document the software developed during the internship and prepare a comprehensive report detailing the results and findings of the investigation.