智能化指挥控制视角下的博弈强化学习

Game reinforcement learning from the perspective of intelligent command and control

人工智能技术的快速发展使得各国信息化建设水平不断提高。智能化是未来战争形态发展的必然趋势,智能化战争对指挥控制也提出了新的要求,即智能化指挥控制。通过对智能化指挥控制与传统指挥控制的分析发现,智能化指挥控制需要博弈强化学习等人工智能技术为其提供支撑。博弈强化学习能够提升指挥控制中态势预测、方案评估和行动控制等能力,但博弈强化学习仍有诸多瓶颈问题。为进一步加深和拓宽博弈强化学习的理论研究,给出了基于图论、元均衡、复因子动力学和分数阶微积分等不同领域的新型解决思路。最后,为验证上述理论提出了基于演化博弈的EG-Q-Learning强化学习算法模型,并阐述了关于当前智能技术的几点思考。

The rapid development of artificial intelligence technology is continuously improving the level of informatization construction in various countries.Intelligence is an inevitable trend in the development of future forms of warfare,and intelligent warfare has also put forward new requirements for command and control,that is,intelligent command and control.An analysis of intelligent command and control and traditional command and con‐trol found that artificial intelligence technology,such as game reinforcement learning is needed to provide support to intelligent command and control.Game reinforcement learning can improve situation prediction,plan evaluation,and action control.However,game reinforcement learning has also many bottlenecks.In order to further deepen and broaden theoretical research on game reinforcement learning,new solutions based on different fields,such as graph theory,meta-equilibrium,complex factor dynamics,and fractional calculus are given.Finally,in order to veri‐fy the above theory,the EG-Q-Learning reinforcement learning algorithm model based on an evolutionary game is proposed,and some thoughts on current intelligent technology are elaborated.