Traditionally, human knowledge is represented mainly in form of written natural language, symbolic formulas and pictures. The project ACKREP aims to supplement these representation forms with additional formal (i.e. machine interpretable) representations. The scope of this project is control theory and control engineering (subsumed under "automatic control").
This discipline is characterized by a wide range of application fields from different physical domains and by a heterogeneous methodical landscape from different areas of mathematics, engineering and computer science.
Basically, the motivation for ACKREP is to facilitate knowledge transfer both
How best to achieve these goals is an open question and subject to ongoing research.
Important notice: Our current attempts of formal knowledge representation are still very experimental. We highly encourage anybody interested in this topic to contact the ackrep-team with questions, feedback and/or contribution ideas.
Large parts of control knowledge are comprised of algorithms and are available as executable code, e.g. feedback design, trajectory planning, observability checking etc. However, often this code is not published or at least hard to execute on different environments. This clearly hinders reproducibility and thus knowledge transfer. In [1.1] we therefore propose ACKREP Code (originally just named "ACKRep"): A Git repository which holds control-related code in a special structure (problems, solutions, methods, environment specifications, ...) plus a webservice which checks the solution-entities against the problem-entities, i.e. a specialized continuous integration service.
In [2.1] we propose the Methodnet, a supplement to classical knowledge representation, consisting of types and methods in a graph structure. From that, a schematic solution procedure can be generated for a specific problem. "Trajectory tracking control for a triple pendulum" is used to demonstrate how the proposed method supports knowledge transfer. Furthermore, an OWL-ontology is automatically generated from the methodnet which allows accessing this knowledge base with SPARQL and other semantic methods.
Note: Currently, information about this project is only available in German.