Digital Twin Framework And Application Based On RoboCup Simulation 2D

RoboCup means Robot World Cup,which is a large-scale and influential international academic competition.RoboCup simulation 2D has become one of the ideal platforms for the study of artificial intelligence theory and related algorithms.At present,many team binaries are optimized and improved on the basis of open source.The team's optimization effect will directly determine its competition results.Taking the broader Agent2 D as an example,with more than 50,000 lines of code and huge state space,it is more and more difficult for developers to optimize it in depth with years of improvement and development.In order to solve this problem,based on the optimization work of YuShan Team of Anhui University of Technology in recent years,based on the idea of "Digital Twin",the theory of digital twin framework under the simulation 2D platform is proposed.This framework is called Simulation 2D Digital Twin Model Framework.It has achieved good optimization results.The main contents of this paper are as follows:Firstly,the development course of RoboCup simulation 2D,the mechanism of competition platform and the general situation and optimization research of teams at home and abroad are briefly introduced,and the binary optimization work of YuShan team of Anhui University of Technology is introduced,analyzed and summarized in detail.The digital twin framework theory of simulation 2D platform is proposed.Secondly,the connotation of the digital twin framework of the simulation 2D platform is elaborated.Building a framework based on physical layer,information layer and digital twin layer.Physical layer saves the game log files such as RCG and RCL produced by binary through management and execution of team binaries.The information layer further extracts the players' basic data through log files,and stores and manages them in XML format.The digital twin layer uses machine learning algorithm to mine players' basic data,and finally establishes players' portraits.At the same time,the feedback system in this layer feeds back the digital features of the players' portraits to the physical layer,thus providing guidance for the optimization of binary.Finally,guided by the digital twin framework of the simulation 2D platform,this paper studies the solution to the optimization problem of team attacking and defensive judgment,finds out the key factors of optimizing attacking and defensive judgment,and points out the optimal way of binary for developers.In order to evaluate the method,the digital features in the players' portraits are processed into data sets,and the Naive Bayesian Classifier is used to classify the players.The accuracy is obtained,and it is stable at about 85%.The practicability of the optimization method and the effectiveness of the guidance of the digital twin framework of the simulation 2D platform are verified.