Early Safety Assessment of Automotive Systems Using Sabotage Simulation-Based Fault Injection Framework

As road vehicles increase their autonomy and the driver reduces his role in the control loop, novel challenges on dependability assessment arise. Model-based design combined with a simulation-based fault injection technique and a virtual vehicle poses as a promising solution for an early safety assessment of automotive systems. To start with, the design, where no safety was considered, is stimulated with a set of fault injection simulations (fault forecasting). By doing so, safety strategies can be evaluated during early development phases estimating the relationship of an individual failure to the degree of misbehaviour on vehicle level. After having decided the most suitable safety concept, a second set of fault injection experiments is used to perform an early safety validation of the chosen architecture. This double-step process avoids late redesigns, leading to significant cost and time savings. This paper presents a simulation-based fault injection approach aimed at finding acceptable safety properties for model-based design of automotive systems. We focus on instrumenting the use of this technique to obtain fault effects and the maximum response time of a system before a hazardous event occurs. Through these tangible outcomes, safety concepts and mechanisms can be more accurately dimensioned. In this work, a prototype tool called Sabotage has been developed to set up, configure, execute and analyse the simulation results. The feasibility of this method is demonstrated by applying it to a Lateral Control system.The authors have partially received funding from the ECSEL JU AMASS project under H2020 grant agreement No 692474, the UnCoVerCPS project under H2020 grant agreement No 643921 and MINETUR (Spain)