A learned soccer goalkeeper Petri net model

摘要

This paper proposes a new learning-based behavior model for a soccer goalkeeper robot using Petri nets. The work aim at modeling and analyzing, both qualitatively and quantitatively, the goalkeeper role in order to have a model-based knowledge of the task performance in different possible situations. The different primitive actions and behaviors as well as the events to switch between them, and also environment models were designed and implemented. For this purpose, a modeling and analysis framework based on Petri nets is used, which enables modeling a robot task, analyzing its qualitative and quantitative properties and using the Petri net representation for actual plan execution. The proposed model building blocks and some tasks of robot are detailed. The novelty of approach is considering some alternatives through tasks execution, which are implemented by conflicts in their Petri net models, and also Q_learning employment in these decision points in order to learn the best policy. Therefore, the execution of actions in different tasks will be controlled effectively. Results of theoretical analysis of some case studies show impressive performance improvement in goalkeeper task execution.