Skip to content

Coffee Machine Example

This example shows how to train a model to predict the behavior of an automaton. Inspired by the work of Steffen et al.1, we consider a simple coffee machine. LearnLib2, a framework for automata learning written in Java, is used for the inference of the automaton.

Todo

This explanation is outdated. The automaton example has to be refined using time series data.

Run this example

Docker Container

The external learner written in Java is containerized for this example. Thus, Docker must be installed to run it. Follow the instructions on the official Docker page to install it. The image for this example is stored in the TUHH image registry. Authentication is required for this. Log in to the registry using docker login collaborating.tuhh.de and provide your credentials when asked. Afterwards run the run.py for this example.

The required images will be automatically retrieved, and a container will be started.

Local Execution

It is also possible to directly run the Java server locally. Before running the example in Flowcean, the server-side has to be started manually by running the Java-project in java/AutomataLearner. A JDK and the Maven build automaton tool are required. More information on Maven projects using the LearnLib library can be found on their website. It has to be assured that the ip and port match those of the server configuration in Java.

  1. Bernhard Steffen, Falk Howar, and Maik Merten. Introduction to Active Automata Learning from a Practical Perspective, pages 256–296. Springer Berlin Heidelberg, Berlin, Heidelberg, 2011. doi:10.1007/978-3-642-21455-4_8. ↩

  2. Malte Isberner, Falk Howar, and Bernhard Steffen. The open-source learnlib - A framework for active automata learning. In Computer Aided Verification (CAV), volume 9206 of Lecture Notes in Computer Science, 487–495. Springer, 2015. doi:10.1007/978-3-319-21690-4_32. ↩