A Model-based diagnosis integrated architecture for dependable autonomous robots

Robots operating with high levels of decisional autonomy require robust fault management architectures capable of effectively handling component faults occurring at various levels of abstraction within the robotic system. In this study, we propose a diagnostic module designed to address dependability concerns in autonomous robots. Our architecture integrates model-based quantitative (residuals-based) and qualitative (logic-based) techniques. To demonstrate the feasibility of our approach, we employ modeling and computer simulation. The simulation incorporates fault injection scenarios to evaluate their impact on the robot’s trajectory during navigation. We conduct a detailed analysis of a testable subsystem of the vehicle, showcasing accurate diagnoses achieved through the integrated processing of residuals for fault detection. Additionally, we
employ automated reasoning using an Answer Set Programming (ASP) engine to achieve fault isolation.