Model-based Robot Design in Modelica

Abstract

Modelica is widely used to support model-based development by enabling engineers to model acausal connections and apply object-oriented design features in the modelling process. Parts of the system are then encapsulated in sub-models, thus avoiding the complexities of a flat model and promoting model reuse. For these reasons we are investigating the uses of Modelica in model-driven robot design.
One of the key goals of our work is to have a system that supports multiple robot configurations without requiring high effort and duplicate work during modelling. In order to achieve this, we developed a simple modular framework, which enables the easy replacement of sub-models through the use of the redeclaration mechanism in Modelica. In this framework models can be created through composition of other models, but instead of specific models they only use abstractions. Components can be then defined independently, and models can be easily configured to use different implementations. This way the components of the robot form a hierarchical tree structure from the concept of the robot itself down to the motors and controllers, where every component has an abstract declaration describing its interfaces, and multiple interchangeable implementations. In addition to this, a common data handling practice across the whole framework further enables easy parametrization of models and data export. Relevant parameters of the component are loaded from a separate Modelica record, so parameter sets are easy to change.
We are currently focusing on serial manipulators and used this framework to model an LBR iiwa robot arm with different levels of detail. The framework is showing great promise for our uses with regard to flexibility and extensibility. The next steps are to test the system on multiple robots, run simulations with robot controller attached and automate data export and import. In this presentation we would like to show the framework design and the Modelica tools utilized on the example of the LBR iiwa model, as well as to discuss the possible uses of our approach in model-driven design.