论文部分内容阅读
在充满竞争的环境中,资源有限导致智能体之间存在利益冲突,有必要建立对手模型并对其行为进行准确预测,从而制定对自身有利的策略.利用交互式动态影响图对未知对手进行建模,将对手的候选模型保存在模型节点并随时间更新其信度.结合观测到的对手动作,在模型空间中利用“观察-动作”序列逐步排除候选模型,最终判定对手的真实模型.实验结果表明,所提出的算法取得了很好的效果,验证了该算法的实用性.
In a competitive environment, limited resources lead to conflicts of interest among the agents, and it is necessary to establish an adversary model and accurately predict its behavior so as to formulate a strategy that is beneficial to itself.Using an interactive dynamic impact graph to build an unknown opponent The candidate model of the opponent is saved in the model node and the reliability of the model is updated over time.According to the observed opponent’s action, the candidate model is eliminated step by step in the model space with the “observation-action” sequence, and finally the opponent’s real model The experimental results show that the proposed algorithm achieves good results and verifies the practicability of the algorithm.