Design-time safety assessment of robotic systems using fault injection simulation in a model-driven approach

The rapid advancement of autonomy in robotic systems together with the increasing interaction with humans in shared workspaces (e.g. collaborative robots), raises pressing concerns about system safety. In recent years, the need of model-driven approaches for safety analysis during the design stage has gained a lot of attention. In this context, simulation-based fault injection combined with a virtual robot is a promising practice to complement traditional safety analysis. Fault injection is used to identify the potential safety hazard scenarios and to evaluate the controller's robustness to certain faults. Besides, it enables a quantitative assessment w.r.t. other techniques that only give qualitative hints, such as FMEA. Thus, it facilitates the refinement of safety requirements and the conception of concrete mitigation actions. This paper presents a tool-supported approach that leverages models and simulation-assisted fault injection to assess safety and reliability of robotic systems in the early phases of design. The feasibility of this method is demonstrated by applying it to the design of a real-time cartesian impedance control system in torque mode as a use case scenario.