Vers des robots collaboratifs autonomes sûrs de fonctionnement

摘要

This manuscript of HDR (Habilitation à Diriger des Recherches, french accreditation to supervise research) presents research work of Jérémie Guiochet carried out at LAAS-CNRS in the Dependable computing and Fault Tolerance (TSF) team. His research work is mainly related to the dependability of collaborative autonomous robotic systems. Specific challenges raised by these systems, including human-system physical interactions and the presence of uncertainties in the perception and decision mechanisms, induce the need to revisit dependability and risk analysis methods. The main contributions address the following topics: safety assessment of collaborative robotic systems (hazard Identi- fication with UML-HAZOP and quantitative assessment of confidence in safety cases), and fault tolerance mechanisms for autonomous robotic systems (redundant planning and synthesis of on-line verifiable safety rules). This manuscript also opens perspectives in the fields of testing of autonomous robots in virtual worlds, uncertainty management for the certification of autonomous robots in human environments, and safety monitoring at different levels in an autonomous software architecture.