陆战Agent学习机理模型研究

Research on Land Combat Agent Learning Mechanism Model

陆战Agent是陆军作战复杂系统ABMS核心的基础要素,学习是陆战Agent适应复杂动态陆战环境的重要能力,如何构建符合陆军作战特点的陆战Agent学习机理模型,是陆军作战复杂系统ABMS必须要解决的关键问题之一。通过陆战Agent基于效果学习本质特征和强化学习算法的分析,结合陆战Agent通信和指挥控制的特点,提出了基于知识共享的陆战AgentPS强化学习机理模型。与一般强化学习模型相比,该模型既能解决感知混淆和学习一致性的问题,又能节省存储空间,提高运行效率,还可实现不同形式的知识共享,增强陆战Agent系统的整体学习和完成作战任务的能力。

The Land Combat Agent＇（LCA） is the core of land combat complex system＇ABMS＇s basics. Learning is the important ability to adapt the complex and dynamic land combat environment. How to construct the LCA learning mechanism model that accords with the traits of land combat is the key to land combat complex system ABMS. Propose the knowledge-sharing-based PS reinforcement learning mechanism model of the LCA by reinforcement learning arithmetic analysis and analyzing the essential traits of effect-based learning of the LCA and the characteristics of C3（communication, command and control）.Compared with generic reinforcement learning models, this model can solve the problems of perceptual aliasing and concurrent learning, save the memory space, increase the run efficiency, realize multiform knowledge sharing and enhance the whole abilities of learning and combat of the LCA.