基于CNN-BiLSTM网络模型的无人机飞行质量评价  

UAV Flight Quality Evaluation Based on CNN-BiLSTM Network Model

作　　者：罗晶 高永 梁葆华[3] 刘军民 惠永昌 LUO Jing;GAO Yong;LIANG Baohua;LIU Junmin;HUI Yongchang(School of Mathematics and Statistics,Xi'an Jiaotong University,Xi'an 710049;College of Aeronautical Basic,Naval Aviation University,Yantai 264001;Chinese Flight Test Establishment,Xi'an 710089)

机构地区：[1]西安交通大学数学与统计学院,西安710049 [2]海军航空大学航空基础学院,烟台264001 [3]中国飞行试验研究院,西安710089

出　　处：《工程数学学报》2023年第2期171-189,共19页Chinese Journal of Engineering Mathematics

基　　金：国家自然科学基金(61877049);科技创新计划(201809164CX5JC6).

摘　　要：为了更好地挖掘无人机飞行轨迹数据中蕴含的有效信息,准确客观地基于轨迹数据对无人机飞行质量进行评价,提出一种融合卷积神经网络(CNN)和双向(Bi-directional)长短期记忆(LSTM)神经网络的CNN-BiLSTM网络模型。首先,利用CNN网络和BiLSTM网络分别获取飞行轨迹数据的局部卷积特征和时间特征。然后,将两种特征送入特征融合层,使用融合后的特征进行分类并获得评分标签。针对六个数据集的数值实验表明,模型不仅取得了较好的分类效果,而且具有很好的泛化能力。In order to better mine the effective information contained in the UAV flight path data and evaluate the UAV flight quality accurately and objectively based on the trajectory data,a CNN-BiLSTM network model which integrates convolutional neural network(CNN)and bi-directional long short-term memory(LSTM)neural network is proposed.Firstly,CNN and BiLSTM are applied to obtain the local convolution feature and time feature of flight path data respectively,and then the two features are sent to the feature fusion layer,and the fused features are used to classify and get the rating label.The numerical experiments on six datasets show that the model not only achieves good classification results,but also has good generalization ability.

关 键 词：无人机质量评价 卷积神经网络 双向长短期记忆网络 

分 类 号：V279[航空宇航科学与技术—飞行器设计]