SmartDG: Modelling System-Level Requirements for Component-Based Robotic Software Systems

System-level requirements and non-functional properties pose a significant modelling challenge in software-intensive robotic systems where the software components are developed by different vendors. It is often a manual and empirically non-verifiable proposition to check all non-functional claims made by a software component and to compose a system with confidence. System-level observance of non-functional properties is key to establish their compliance with system-level requirements, which at present is mostly solely a subjective judgement of the system builder. A crucial next step in system composition policy should be to enforce encapsulation of system-level properties within the functional body of the component. A software component should be capable of informing its peers how system-level properties are modified across services that it provides. Tracking system-level properties across components is a desired mechanism to verify compliance of system-level requirements. Monitoring these properties is desired during (i) component development for debugging, (ii) system composition for adequate component selection, and (iii) at run-time for diagnosis and enforcement. This paper presents Smart Dependency Graphs “SmartDG”, an addition to the RobMoSys compliant SmartMDSD toolchain, to model system-level dependencies between robotic software components. In SmartDG methodology, component developers implement transfer functions for dependency objects defined by domain experts, which enables system builders to trace system-level properties via dependency graphs. SmartDG provides a structural template to model system-level requirements to allow for an empirically verifiable composition of a robotic software system.