| With the rise of the new technology, the research of RoboCup soccer simulation 2D has new directions, for example, analyzing log files that are generated from RoboCup soccer simulation platform by the new technology of machine learning and data mining for helping team decision. The paper chooses team formation of RoboCup soccer simulation as a research point, and put forward a team formation recognition algorithm based on density peaks clustering by taking example from the paper published on Science in 2014, the clustering by fast search and find of density peaks. The main works in this paper are following:First of all, the paper gives a brief introduction on RoboCup soccer simulation 2D and its platform, the feature and research direction of each team at home and abroad will also be given detailedly, and this part focus on introducing the research progress of team formation recognition.Secondly, the paper presents the related clustering knowledge, and plays more emphases on machine learning, clustering analysis and the similarity measure method of data objects, summarizes the advantages and features of state-of-the-art clustering algorithms, thus, leads to the clustering algorithms used in the paper. This part gives more details on the thought of clustering by fast search and find density peaks, and its application field, besides, the paper also explores the feasibility of this clustering algorithm on the research field of RoboCup simulation.Finally, the paper extracts and disposals the data of team formation from 2D platform's log files to calculate correlative distance matrix and uses the density peaks clustering algorithm to implement the simulation experiment of team recognition. The result of experiment shows the algorithm is effective for team's formation recognition.The paper mainly applies the new clustering algorithm into the team formation recognition research of RoboCup soccer simulation 2D, and recognizes formation from the related offline log files. |
PDF Full Text Request