群体中Agent基于内部状态的行为选择  

Behavior Selection based on Internal-state for Agent in Group

作　　者：江道平[1] 尹怡欣[1] 班晓娟[1] 孟祥嵩[1] 

机构地区：[1]北京科技大学信息工程学院,北京100083

出　　处：《系统仿真学报》2009年第1期121-126,共6页Journal of System Simulation

基　　金：国家自然科学基金(#60503024;#60374032)

摘　　要：以自然鱼群为原型,研究它们的行为选择机制,提出了一种群体环境中基于内部状态的Agent行为选择方法。建立分布式行为模型,赋予Agent感知、交互能力;用R-A模型描述Agent之间的相互作用;定义Agent的自然能力限制为硬约束,定义Agent间相互作用为软约束;采用基于约束满足的随机搜索算法模拟Agent的交互与行为选择,运用Breakout思想确保算法跳出局部最小;通过对Agent内部状态值的调节改变搜索参数,达到内部状态控制行为选择的目的。To simulate the behavior selection for fish-group, a new way to construct the behavior selection for Agent in group based on its internal-state was proposed. A distributed behavior-model was established, in which the Agent was endowed with abilities of perception and interaction. The R-A model was adopted to define the mutual function between Agents. The limitations of Agent’s perception and motion ability were regarded as Hard-Constrains, and in the same way, the mutual functions of Agents in group were regarded as Soft-Constrains. A random search algorithm based on constrains satisfaction was proposed to simulate the behavior selection and interaction of Agent, in which the idea of ＂Breakout＂ was adopted to guarantee the algorithm to escape from local minimum. Besides, the internal-state values were used to modify constrains-weights in the algorithm, thus the behavior of Agent could be controlled by its internal-state.

关 键 词：群体 Agent内部状态 行为选择 约束 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]