Football Simulation Modeling with Fuzzy Rule Interpolation-based Fuzzy Automaton

Abstract

A Fuzzy Rule Interpolation-based (FRI) fuzzy automaton for controlling a football match simulation is going to be introduced in this paper. Controlling the agents (football players) of the simulation is realized by evaluating such fuzzy rule-bases, which contain only the cardinal rules able to make the system work, keeping the rule-bases as small as possible (forming so called sparse rule-bases). Classical fuzzy inference methods require complete rule-bases by design and cannot handle these kinds of sparse rule-bases. However, using sparse rule-bases to control the agents becomes possible by applying FRI. The goal of this work was to construct such a model, which employs a human-readable knowledge representation to control the agents in a football simulation. For this purpose, the application of sparse fuzzy rule-bases is well suited, as these are self-describing by their nature. An example application was also developed alongside the fuzzy automaton-based model, which is able to perform and show a lifelike football match simulation in real-time. Hence the presented model can be adapted to real robot hardware and also can be used as a reference model for fuzzy logic based machine learning methods.