Orchestration Graph: From Knowledge Graphs to Executable Analysis Workflows with openCAESAR
TL;DR
The Orchestration Graph is introduced, an approach that transforms ontological models into executable analysis structures, enabling knowledge graphs to directly drive workflow execution through bindings to external computational functions.
Abstract
Knowledge graphs are widely used in Model-Based Systems Engineering (MBSE) to represent systems engineering knowledge and support semantic reasoning and interoperability. However, integrating semantic knowledge graphs with mathematical analysis and simulation remains a key challenge. In practice, multi-domain analysis workflows are typically constructed manually using scripts. This paper introduces the Orchestration Graph, an approach that transforms ontological models into executable analysis structures, enabling knowledge graphs to directly drive workflow execution through bindings to external computational functions. Using the Ontological Modeling Language (OML), system variables and constraints are formally defined, while their dependencies are represented as a directed acyclic graph. The graph is further extended by binding model elements to external computational functions, such as Python and R, enabling automatic workflow synthesis, function invocation, and parameter propagation. The resulting Orchestration Graph derives execution logic directly from the model, allowing automated synthesis and execution of analysis workflows. The approach is demonstrated through a ΔV budget analysis that estimates the velocity changes required for spacecraft maneuvers and the associated propellant mass. Furthermore, the proposed methodology is extended to support design space exploration. Results indicate that the approach enables reusable and automated analysis while supporting continuous verification and validation from the early stages of system design.