Mining And Application Of Robocup Simulation 2D Game Log

Multi-agent system, is one of the hottest issues in the field of artificial intelligence research. Multi-agent system, can simulate human intelligent behavior of labor cooperation, and promote the development of multi-agent research field by using all kinds of artificial intelligence theory and combined with the real world. RoboCup which is a worldwide competition and academic activities is created to promote the development of the research of distributed robotics. As the new standard question of artificial intelligence and robot, RoboCup, having many characteristics of multi-agent system, is a good experimental platform for all kinds of new theory and technology. Among them, RoboCup 2D, based on multi-agent decision-making competition key, provides an opportunity in all over the world that is these scholars of multi-agent research field could communicate with each other.This paper is trying to extract some data from the log files of RoboCup simulation 2D competition and analyze some results to improve the winning percentage of RoboCup simulation 2D competition by using the technology of data analysis which is wild applied in the Internet field now.The log files of RoboCup simulation 2D competition records the position, speed of the players and the ball, the players’physical strength value, view and all sorts of properties in all cycle of game, it includes almost all the details of the game. Therefore, to directly check out his team’s shortcomings, the competitors are usually replaying the game video by calling the log files, and then to modify the corresponding code. This is a most common way that the log files being used.In the RoboCup simulation 2D competition, dribbling from the self-area to the penalty area of opponent is the best way to show the attack force of a team. Because the only way to get scores is ours player dribble to the opponent’s penalty area, and in competition, everyone would trying to block this action. Therefore in this situation, the most common strategy is to intercept or slide tackle to block the attack of opponent’s player. And in attacking process, which is the attacker attempt to dribble into defender’s penalty area, the most common strategy is including to the ball handler’s and non-ball handler’s, to the ball handler, the best strategy is including Dribble, Pass, Hold, and to the non-ball handler, the best strategy is running position.This paper is trying to extract some data from the process of the breakthrough of midfielder, to analyze the ball handler’s strategy, and also to statistics the number of times of successful breakthrough the defense. Then by analyze the correlation of all kind of factors and using the collaborative filtering Algorithm, we could obtain some best strategy of ball handler from this method. In the end, relying on the final analysis result we will modify program to improve our team’s performance, and verifying this method is reasonable and feasible.