The Application of Hybrid Distributed Reinforcement Learning Algorithm in RoboCup 2D Soccer Simulation System

Abstract

AbstractThe RobuCup 2D Soccer Simulation has been used as the basis for successful international competitions and research challenges.When developing a soccer team for simulated soccer ,one has to deal with the RoboCup soccer simulator. In this simulation system, it is the Multiagents’ sensory perceptions and individual abilities which are received from the server, coupled with the cooperation strategies designed by researchers, that together determine whether the soccer team will win this game or not. In cooperation strategy designing, researchers often employ the method of Distributed Reinforcement Learning algorithms to optimize the simulation system. The paper is extended as follows: First, we will make a description of the characteristics about the RoboCup simulation system. Second, there is an analysis of the limitation about the traditonal application of Distributed Reinforcement Learning Algorithm, and a new solution will be advanced. Finally, Simulated results will be discussed and the paper comes to a conclusion.KeywordsRoboCup 2D Soccer Simulation SystemmHybrid Distributed Reinforcement Learning AlgorithmMultiagent