Application of Neural Networks in Differential Game
-
摘要: 阐述了微分对策理论在实际应用中存在的问题,并从动态冲突的实际过程出发,将神经网络理论与分层式智能控制理论相结合,来模型化微分对策问题.将微分对策的双向优化问题转化成基于神经网络理论的目标机动性辨识和单向优化控制问题,将微分对策的定性与定量结合起来,为有效加入人的经验提供了一种技术途径.并采用自适应分布式学习速率,大大提高了神经网络的学习速度和精度.Abstract: The existing problems of application of differential game in practice are analyzed.Considering the processes of practical dynamic conflicts,neural networks combined with the theory of multilayer intelligence control are introduced to modeling the problem of differential game.It turns the problem of differential game into the problem of maneuver identification and optimization control based on neural networks.This process of solving differential games is more conformity with practice,and provides an efficient method for adding human experience and combining the qualitative analysis with quantitative analysis of differential game.The adaptive distributed learning rate of neural networks is developed to increase the learning speed and performance greatly.
-
[1] 许 成,沈如松,周卿吉.基于神经网络和微分对策理论的制导律[J].系统工程与电子技术,1998,20(1):1~4. [2]蒋映忠.空战微分对策的智能控制研究 .北京:北京航空航天大学自动控制系,1994. [3]Rodin E Y,Lirov Y.Artificial Intelligence in air combat games[J].Computer and Mathematics with Applications,1987,13(3):261~274. [4]Rodin E Y.Semantic control theory[J].Applied Mathematics Letter,1988,1(1):73~78. [5]Rodin E Y,Amin S M.Maneuver prediction in combat via neural networks .Computers Math Applic,1992,24(3):95~112. [6]Harmon M E,Baird L C.Multiplayer residual advantage learning with general function approximation ．U.S. Air Force Academy,1996. -
点击查看大图
计量
- 文章访问数: 2613
- HTML全文浏览量: 226
- PDF下载量: 673
- 被引次数: 0
下载:
