Formation Learning Of RoboCup Simulation 2D Based On Player Cooperative

With the arrival of the machinery industry 4.0 times, the trend of industrial intelligent is unstoppable, robot is an important part of the process, it involves many scientific frontiers. The Robot Soccer World Cup and Conference(RoboCup) is a standard for studying artificial intelligence and robotics, and provides a platform to exchange for reaearchers. There are simulation groups、entity groups、humanoid groups and standard groups in RoboCup.RoboCup simulation 2D is one of the traditional RoboCup competitions, has focused on high-level decision-making, such as, moving withoutball、shooting techniches、ball intercepting and so on.Players’ actions depend on decision-making, which is based on the team formation and has a critical impact on the outcome of the game.Like humans football game,the formation is related to tactical,so it’s important and has a great significance learning opponents’ formation and makes the targeted tacticals.This paper describes the development and research directions of RoboCup2 D home and abroad,and then analyzes the research status and significance of formation learning.Second, RoboCup2 D platform is a distributed, real-time,instability,and limited comunication platform.This paper studies the decision-making in depth problems after describes the mechanism of RoboCup2 D.This paper also analyzes formaiton modeling which based on role, and describes the formation strategies combined with current teams UVA and agent2 d.We mine the moving mode based on the formation, which can provide a foundation of formation learning.Finally, degree of palyers’ relationship is an important concept in multi-agent collaboration, this paper presents degree matrix of players’ relationship as identity of formation. We analyze several key factors that affect degree of players’ relationship, and then do math modeling and derive the calculation of degree between players. We obtain the template formation from log files, and use the template formation to learn the actual team formation.The log files include rcg file and rcl file. We use the offline data from log files to program to make true the algorithm, and then design simulation experiment of formation learning. The methord is used in the development of YuShan2 D which is form Anhui University of Technology. The outcomes certificate its effectiveness in RoboCup World Cup and RoboCup China Open and Competition of Anhui Province. Future work is an online real-time formation learning.