Model-Driven Run-Time Dependency Graphs for Component-Based Robotic Software

For the observance of a system's non-functional properties, one must first parameterise them as elemental or compound objects. Dependency objects are defined as entities that encapsulate a distinct system characteristic that is relevant to and is altered by components and connections that compose the system. Software components and network connections form nodes of a graph, edges of which represent a change of one or more system-wide properties. Dependency graphs are networks of dependencies that emerge from the flow of system-level characteristics, controlled by evidence and business logic. Component developers and network experts provide functional units to describe how the value of a particular dependency object is transferred across computational or connection nodes respectively. Run-time dependency graph models for independently developed software components and network connections are integrated, built and explored while building or validating a system. These run-time models are used to observe networks of dependencies that emerge from the flow of system-level features graphically and programmatically. The design and engineering anatomy of such a toolchain is presented in this paper.