改进决策树的无人机空战态势估计

针对无人机空战态势估计中存在的多参数、非线性、实时性问题,提出了一种改进决策树思想的态势估计推理方法.首先,通过结合无人机与敌机的状态参量作为决策树模型的输入,确保态势估计的依据中包含交战双方的信息,为无人机态势估计的结果的合理性提供理论依据;然后,建立4类态势结果作为决策树模型的输出,以满足态势响应快速性的要求,根据影响空空导弹攻击区的状态参量,对比相同状态参量下,无人机与敌机的评价指标值的大小,构建对应的空战态势分类指标.建立了空战态势分类规则,作为决策树的推理规则,在决策树的节点对态势不断细化.最后,针对决策树中未开发分支引入反推理规则,在未知情形下提高学习能力.通过对不同的典型空战场景:一对一、一对二和二对二,进行仿真验证,并将结果与贝叶斯推理法进行全面比较,通过分析,所提方法用时5.39 s,准确度为80%,贝叶斯推理法用时11.63 s,准确度为60%.准确的实验结果表明所提方法比贝叶斯推理方法的评估速度更快,准确度更高.

Situation assessment for unmanned aerial vehicle air combat based on anti-reasoning rules decision tree

For the multi-parameter, nonlinear, and real-time problem in air combat situation assessment, an approach is proposed based on a novel structure of decision tree. Ensuring the basis for the assessment of the situation includes obtaining the information of the warring parties and the state parameters of the unmanned aerial vehicle(UAV). The enemy was used as the input of the decision tree and four kinds of situation results as the output of the decision tree, which provided a theoretical basis for the rationality of the result of the UAV situation estimation. According to the state parameters that affect the attack area of the air-to-air missile, by comparing the same state parameters, the magnitude of the evaluation index value of the UAV and the enemy aircraft were obtained, and the corresponding indicators of the air combat situation classification were designed to meet the rapid response requirements of the situation assessment. The rules of the air combat situation classification were established to be inference rules of the decision tree. Finally, anti-reasoning rules were proposed for undeveloped branches in the decision tree in order to improve learning under unknown conditions. Extensive simulations, including one-to-one, one-to-two, and two-to-two air combat scenarios, show that the computing time of the proposed method was 5.39 s and its accuracy was 80%, while the computing time of the Bayesian was 11.63 s and the accuracy was 60%. The results indicate that the proposed optimal method has faster assessment speed and higher accuracy than the traditional Bayesian network.