RoboCupSoccer中多机器人协作截球策略  

Collaboration interception strategy of multi robots in RoboCupSoccer

作　　者：蔡剑怀[1,2] 李茂青[1] 吴顺祥[1] 

机构地区：[1]厦门大学信息科学与技术学院,福建厦门361005 [2]厦门大学生命科学学院,福建厦门361005

出　　处：《计算机工程与应用》2010年第16期1-5,62,共6页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60704042);航空科学基金(No.20080768004);厦门大学985二期信息创新平台项目~~

摘　　要：截球策略是决定机器人足球队比赛能力的重要因素。由于信息噪音、命令执行误差、多异构类型和随机异构参数等因素的影响,单纯基于个体技术的截球决策并不一定可靠。首先采用数学解析方法建立截球个体技术模型,利用牛顿迭代法求解最快截球周期、截球点和基本命令队列;然后基于BP神经网络描述两个截球周期相比的截球成功概率;最后机器人基于自己和同伴的角色关系、截球成功概率和截球点所在球场区域进行协作截球决策。实验结果显示基于角色的协作截球效果明显改善。基于此策略的机器人足球队在比赛中取得不错的成绩。Interception strategy is an important factor for the competitive ability of robot soccer team.Due to the effect of the information noise,command implementation error,multi-heterogeneous types and random heterogeneous parameters,the interception strategy simply based on individual skill is not necessarily reliable.Firstly,mathematical analysis method is adopted to establish the interception individual skill model,and Newton iteration method is used to get the fastest interception cycle,interception point and the basic command queue.Then BP neural network is adopted to describe the interception success probability comparing two interception cycles.Finally the robot makes the interception decision collaboratively considering the allocation of roles,the interception success probability and the playfield interception point locates.The experiment results show the improvement of the rolebased collaboration interception.The robot soccer team with such strategy gets good results in the competitions.

关 键 词：RoboCupSoccer 截球模型 BP神经网络 角色 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]