基于Q学习的管制员Agent学习行为研究

Research on learning behavior of air traffic controller based on Q-learning

管制员Agent是空中交通运行仿真系统中的核心部分,为了提高其知识库的完备程度,做到空中交通的精确仿真,可以考虑将机器学习理论引入管制员Agent模型.研究了相关机器学习算法,提出管制员Agent的个体机器学习行为,选择Q学习算法对管制员Agent的学习行为进行建模,使管制员Agent能在空中交通运行仿真中取得最优策略,完善自身冲突解脱知识库的不足.仿真结果证明了管制员Agent学习行为的合理性.

ATC Agent is the core part of an air traffic operation simulation system. In order to increase the degree of completeness of its knowledge base and achieve accurate simulation of air traffic, the machine learning theory was introduced into ATC Agent model. This paper studied the relevant machine learning algorithms and presented the individual learning behavior of ATC Agent. Then Q - learning algorithm wasselected to model the learning behavior of ATC Agent. Thus the ATC Agent was able to obtain the optimal strategy in the process of air traffic operation simulation and improve its knowledge base of conflict resolution. The simulation result proved the rationality of learning behavior of the ATC Agent.