模型数据混合驱动的水声器材防御决策方法

Underwater Acoustic Equipment Defense Decision Method Driven by Both Model And Data

基于模型的计算机仿真技术在水中对抗战法推演、方案制定等应用中至关重要。但是由于战场输入数据维数高且模型复杂度高,基于模型仿真方法很难从算法、计算优化的层面上满足决策实时性要求。基于仿真模型产生的大量仿真样本数据,采用学习算法拟合求解最小极值点,缩小决策空间,再与传统仿真模型相结合以提高过程仿真优化效率;实验表明,采用模型与数据驱动混合学习模型在决策相对较优条件下可以将决策时间缩短到原来的6%。在积累了大量作战仿真、训练数据的今天,采用机器学习算法代替传统基于模型算法,是满足实时性和决策准确率的新思路。

Model based computer simulation technology is very important in the application of tactics deduction and scheme formulation in the process of underwater confrontation.However,due to the high dimension of battlefield input data and high model complexity,it is difficult for the existing model-based simulation methods to meet the real-time requirements of decision-making from the aspects of algorithm and calculation optimization.Therefore,this paper considers a large number of simulation sample data generated based on the simulation model.Firstly,the learning algorithm was used to fit and solve the minimum extreme point to reduce the decision space,and then combined with the traditional simulation model to improve the efficiency of process simulation optimization.Experiments show that the hybrid learning algorithm driven by model and metadata can reduce the time of model-based decision-making by 94% under relatively optimal decision-making conditions.With the accumulation of a large number of combat simulation and training data,using machine learning algorithm to replace the traditional model-based algorithm is a new idea to meet the real-time and decision accuracy.