战场对敌目标战术意图智能识别模型研究  

Study on The Intelligent Recognition Model of Enemy Target's Tactical Intention on Battlefield

作　　者：欧微[1,2] 柳少军 贺筱媛[1] 曹占广[1] 

机构地区：[1]国防大学信息作战与指挥训练教研部,北京100091 [2]乌鲁木齐民族干部学院,新疆乌鲁木齐830002

出　　处：《计算机仿真》2017年第9期10-14,19,共6页Computer Simulation

基　　金：国家自然科学基金(61403401;61374179;61273189;61174156;61174035);全军军事学研究生课题(2015JY035)

摘　　要：对战场敌方目标战术意图的快速、准确和自动识别,是智能决策的前提和基础。针对传统意图识别模型在知识表达、网络训练和时序特征学习上面临的困难,提出一种模拟指挥员进行情况判断时的记忆机制和推理模式、基于长短时记忆循环神经网络的战场目标意图智能识别模型,构建了模型的基本框架、设计了相应的时序特征编码方法、标签知识封装与模式解析机制,并通过采用学习因子自适应调整策略提高模型的训练效率。测试结果表明,所提模型具有比传统循环神经网络模型更好的收敛性能,能以较高的识别准确率实现对敌方目标战术意图的自动识别。Automatic and fast intention recognition is the premise and bedrock of intelligent decision-making. Therefore, designing efficient intention recognition models with the ability of learning temporal features are critical. To overcome the difficulties of knowledge expression and temporal features learning appeared in traditional algorithms, an intention recognition model based on recurrent neural networks （RNNs） that possesses long-short term memory is proposed, which simulates the commanders＂ memory mechanism and reasoning mode when they are judging the battle- field situation. Moreover, the basic framework of the intelligent model, the coding method of temporal characteristics, the knowledge encapsulation and pattern parser mechanism are also introduced, meanwhile, a self-adaptive algorithm is proposed to improve the train efficiency of the model. The experimental results show that the proposed model has better convergence performance than the traditional recurrent neural network model, and can automatically recognize the target＇s tactical intention with high recognition accuracy.

关 键 词：意图识别 时序特征 循环神经网络 长短时记忆网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP191.9[自动化与计算机技术—控制科学与工程]