Master theses

Data-driven workflow planning is powerful when processes are linear: each step consumes data and produces data for the next step. But real workflows are rarely linear. They contain control-flow constructs such as conditionals, exclusive choices (XOR branches), loops, and splits/joins that depend on runtime data and execution state. These constructs are standard in workflow systems, yet they remain difficult to model declaratively in knowledge-driven planning pipelines.

The FnO-Steps specification provides a strong RDF-based foundation for function-oriented workflow description, composition, and execution. However, current modeling patterns focus mainly on straight-line compositions. As soon as advanced control flow is needed, developers often fall back to ad-hoc engine-specific logic, which reduces interoperability, makes workflows harder to validate, and limits automatic planning and optimization.

This creates a key research and engineering challenge: how can we represent branching and iteration as first-class, declarative concepts in FnO-Steps, while keeping workflows machine-composable, semantically precise, and executable across implementations? Solving this would significantly increase the expressiveness of semantic workflow planning and make FnO-Steps viable for more realistic orchestration scenarios.

In this thesis, you will design and prototype an extension of FnO-Steps that introduces explicit control-flow primitives such as ConditionalStep, LoopStep, and SplitStep, together with their execution semantics and validation constraints. You will define how these new constructs interact with existing FnO and FnO-Steps concepts, including data dependencies, preconditions, and step outputs.

Concretely, you will propose an ontology-level design for control-flow constructs, implement support in the OSLO Steps Workflow Composer, and demonstrate end-to-end planning/execution on realistic scenarios that require branching and repetition. You will also investigate how these constructs can be compiled or transformed for efficient execution without losing declarative meaning.

You will evaluate your extension on three dimensions: expressiveness (can common workflow patterns be modeled cleanly), correctness (do execution results match formal semantics), and interoperability (can workflows be exchanged and reused without engine-specific glue code). This work builds directly on IDLab's existing research and tooling, including prior publications such as this work, and offers a concrete path toward a more complete semantic workflow standard.

The expected output is a validated extension to core control-flow patterns and a demonstrator in existing tooling, not a full workflow engine rewrite.