Abstract: In the fire strategy of RoboCup rescue simulation, controlling the spread of fire plays an important role in rescue effect. Most teams use single-objective selection method based on various indicators of the single building. It is difficult for this method to control the fire effectively when more buildings are in fire or the fire spreads quickly. This paper proposed a new partitioning method based on density cluster. In this method, all the buildings were clustered and seperated into different regions according to the fire spread speed, the target region and buildings extinguishing the fire were selected by the weight considering both the fire spread speed and the building attribute, thus the fire could be controlled even completely extinguished. Finally, the validity of this method has been confirmed through simulation test and competition.

Key words: rescue simulation, cluster analysis, fire strategy, partitioning

摘要： 在RoboCup救援仿真的消防策略中，火势蔓延的控制对救援的效果有着重要的作用。大多数队伍采用基于单个建筑的各种指标的单目标选择法，该方法在着火建筑较多或者火势蔓延快时难以对火势进行有效控制。提出一种新的基于密度聚类的分区方法，根据火势蔓延速度将所有建筑进行聚类分析形成建筑簇并分开成为不同的区域，然后综合建筑的属性作为权值选择目标区域和灭火建筑，从而有效地控制火灾的蔓延甚至将火灾完全扑灭。最后通过仿真实验和比赛验证了该方法的有效性。

关键词: 救援仿真, 聚类分析, 消防策略, 分区