多种评价指标下RoboCupRescue救援智能体算法

Ambulance team algorithm based on multi-evaluation index in RoboCupRescue

为了促进RoboCupRescue项目的发展,使救援智能体的表现在多种目标向量指标下获得更好的评价,通过分析13个计分向量的变化规律,研究与救援智能体(ambulance team,AT)直接相关的6个计分向量在不同救援策略下的评价惩罚,提出兼顾相关评价指标的AT调度算法.该算法包括运用粒子离散方法对市民的降级时间进行预测、基于评价惩罚规则对市民救援优先度进行排序和根据最有效救援集合对AT进行救援目标的分配,并权衡了各个计分向量相互间的影响,从而优化救援智能体的表现.最后在综合性的灾难环境Foligno-3的测试取得了具有代表性的结果,表明该算法是有效的.

In order to contribute to the progress of RoboCupRescue and make the performance of ambulance team agents obtain better evaluation under the indices of multi-vector points,a new ambulance team(AT) algorithm which considers relevant evaluations is proposed by analyzing the regularity for changes of 13 vector points and studying the evaluation of six vector points directly related to the AT with different rescue strategies.It includes forecasting the downgraded time on the civilians by using the particle discretization method,sorting the rescue priority of civilians based on the rules of evaluation of penalty and distributing the rescue targets among ATs according to the most effective aid collection.Moreover,it balances the mutual effects between each vector point so as to optimize the performance of ambulance team agents.Finally,the test in Foligno-3 disaster environment proves the effectiveness of the algorithm.