SACC - A property driven approach to expose undesired behaviors among system’s components

In recent years, there has been an increase in automation of safety critical systems such as self-driving cars, caretaking robots, or rescue drones. With increase in automation, the risk of systems behaving in an undesired manner has also risen. Most safety analysis approaches predominantly concentrate on identifying or assessing how the failure of a component can affect the behavior of a system based on the system’s requirements. Yet, undesired behaviors can also occur when multiple components concurrently exert opposite effects on a shared resource. To identify safety-critical issues due to undesired concurrent component behaviors, we propose a property-driven approach called safety assessment for concurrent components (SACC). SACC uses a combinatorial technique that considers the requirements specification of a system, expressed as the states and properties of the system’s components for identifying undesired combinations of component behaviors. To evaluate SACC, we performed a study using a requirements document on a caretaking robot. Our results show that SACC identified 38%-80% more undesired system behaviors when compared to the control techniques.