Hypergraph-based Modeling of Cognitive Dataflow Systems

Cognitive systems have become widespread in various computer-based applications, from assistants to robotic systems. Although such systems are useful even in their current version, there is a growing need to incorporate semantic information. This paper introduces HyMeKo, a framework for modeling information using hypergraphs, which allows for the creation of multiple relationships between information entities through a LALR(1)-based language. The framework’s primary application contexts are in robotics, cyber-physical systems, and intelligent systems. A key innovation of this work is a transformation method between hypergraph semantic descriptions and e-adjacency tensors. An additional practical result from the implementation of the framework is a hypergraph-based dataflow system, enabling the simultaneous description and execution of dataflow processes. The paper includes a minimal example to compare the compactness of the introduced language with the widespread URDF, to describe a minimal, 2-link system, demonstrating its advantages in terms of conciseness and flexibility.