月表陨坑检测轻量化深度学习方法

针对目前基于深度学习的陨坑检测方法存在的模型参数量大和检测速度慢的问题，提出了一种轻量化的深度学习陨坑检测方法。首先，采用通道剪枝方法删减卷积神经网络中冗余的卷积核，得到结构紧凑高效的陨坑检测模型。然后，使用轻量化的深度可分离卷积操作替换基础陨坑检测模型中的标准卷积操作，进一步降低了模型的复杂度。仿真实验结果表明，所提出的轻量化陨坑检测模型能够保证较高的像素预测精度，并且能够适应亮度、图像噪声等干扰因素的影响。同时，与轻量化处理前的模型相比，参数量减少了99.2%，检测速度提升了94%。     

机动发射弹道导弹集群诸元快速规划

针对机动发射条件下弹道导弹集群的飞行诸元快速规划问题，将神经网络预测与最小二乘优化相结合，提出了一种弹道导弹发射诸元快速规划方法。首先分析了弹道导弹助推段飞行策略并选取适当的发射诸元，以发落点信息为输入，设计双隐藏层诸元预测网络，通过弹道仿真获取弹道数据建立数据集完成网络训练，利用该网络可以得到发射诸元迭代初值。在此基础上，为了消除数据集中样本数据不平衡对发射诸元规划精度的影响，以落点射程、横程、高程偏差最小为指标函数，结合最小二乘优化方法进行迭代获得发射诸元精确解。最后在典型发射场景下，进行了弹道导弹集群机动快速发射仿真验证。结果表明，该方法相较于传统方法可显著提高计算速度与精度，且在给定的大范围机动条件下，能够满足弹道导弹集群对远距离、多目标的快速精确打击。     

Performance improvement of licklider transmission protocol in complex deep space networks based on parameter optimization

A reasonable parameter configuration helps improve the data transmission performance of the Licklider Transmission Protocol (LTP). Previous research has focused mainly on parameter optimization for LTP in simplified scenarios with one to two hops or multihop scenarios with a custody mechanism of the Bundle Protocol (BP). However, the research results are not applicable to communications in Complex Deep Space Networks (CDSNs) without the custody mechanism of BP that are more suitable for deep space communications with LTP. In this paper, we propose a model of file delivery time for LTP in CDSNs. Based on the model, we propose a Parameter Optimization Design Algorithm for LTP (LTP-PODA) of configuring reasonable parameters for LTP. The results show that the accuracy of the proposed model is at least 6.47% higher than that of the previously established models based on simple scenarios, and the proposed model is more suitable for CDSNs. Moreover, the LTP parameters are optimized by the LTP-PODA algorithm to obtain an optimization plan. Configuring the optimization plan for LTP improves the protocol transmission performance by at least 18.77% compared with configuring the other parameter configuration plans.     

运载火箭遥测数据智能判读算法应用

在运载火箭高发射密度、高判读需求、高数据量的背景下，现有自动化判读的判据覆盖率不全、判据编写门槛高、耗时多的问题日益凸显，缺少较通用的算法对传统判读算法未覆盖的判读任务进行判读补充，进而影响运载火箭效果评估与系统性能评定。为充分挖掘海量遥测数据中隐含的参数变化规律，设计智能判读算法作为传统算法的有益补充，提升传统判读的判读覆盖率和判读效率。以液体运载火箭长期加电试验产生的遥测数据为研究对象，设计集成神经网络智能判读算法，在给出的判读指标下研究得出，集成神经网络在频率异常、丢帧等五种现有判据难以描述的判读场景下，判读性能提升30%，提高了现有判据的覆盖率，后续可为判读体系完善和智能判读落地提供研究参考。     

Green UAV communications for 6G: A survey

Unmanned Aerial Vehicles (UAVs) have received a wide range of attention for military and commercial applications. Enhanced with communication capability, UAVs are considered to play important roles in the Sixth Generation (6G) networks due to their low cost and flexible deployment. 6G is supposed to be an all-coverage network to provide ubiquitous connections for space, air, ground and underwater. UAVs are able to provide air-borne wireless coverage flexibly, serving as aerial base stations for ground users, as relays to connect isolated nodes, or as mobile users in cellular networks. However, the onboard energy of small UAVs is extremely limited. Thus, UAVs can be only deployed to establish wireless links temporarily. Prolonging the lifetime and developing green UAV communication with low power consumption becomes a critical challenge. In this article, a comprehensive survey on green UAV communications for 6G is carried out. Specifically, the typical UAVs and their energy consumption models are introduced. Then, the typical trends of green UAV communications are provided. In addition, the typical applications of UAVs and their green designs are discussed. Finally, several promising techniques and open research issues are also pointed out.     

基于极化神经网络的雷达舰船检测识别方法

相参雷达捕获的全极化海面目标距离-多普勒(RD)回波数据中，目标区域占比小、信噪比低，且海况环境与干扰种类多变，使得经典的深度神经网络在此种条件下检测识别精度较低。为此，本文提出了一种基于极化深度神经网络的全极化相参雷达海面目标检测识别算法。首先，引入极化特征提取模块挖掘目标与干扰的差异化特征;其次，通过特征金字塔网络解决小目标检测识别的问题;最后，使用级联结构进一步提升算法性能。在全极化相参雷达回波数据集上的测试结果表明:基于特征值与特征矢量的极化特征对于数据集中两类舰船目标的平均精度分别达到0.907 9与1.0，相比不采用极化特征有着显著提高。     

飞行器制导控制一体化编队控制器设计

针对主从式结构飞行器协同编队控制问题，以侧滑转弯飞行器为研究对象，采用制导控制一体化(Integrated guidance and control, IGC)方法设计编队控制器。首先在惯性坐标系中定义相对运动坐标系，建立相对运动模型，结合飞行器动力学模型，得到全状态制导控制一体化模型；然后采用反演方法，结合滑模变结构与神经网络自适应理论设计了编队控制器，并证明了控制系统稳定性；最后在高速情况下进行了六自由度数值仿真，对比了IGC设计方法与分离设计方法的控制性能。仿真结果表明所设计的IGC控制器能够快速精确地对期望编队队形进行构建与保持，并且较分离设计方法具有优越性。     

增强神经网络辨识模型泛化能力的研究

神经网络(Artificial Neural Network,ANN)辨识模型的泛化能力是其最主要的性能之一,增强ANN模型的泛化能力也是近年来国内外有关专家学者研究的重点问题。大量研究表明,ANN模型泛化能力的改善与很多因素相关联,其中恰当的性能指标函数设计是一个重要影响因素。文中在分析常见的基于均方误差最小原则的性能指标函数基础上,通过加入ANN辨识模型权值间的延迟信息,进而获得一种改进型性能指标函数。通过仿真,验证了所设计的改进型性能指标函数对增强ANN辨识模型的泛化能力是有效的。     

Reinforcement learning based dynamic distributed routing scheme for mega LEO satellite networks

Recently, mega Low Earth Orbit (LEO) Satellite Network (LSN) systems have gained more and more attention due to low latency, broadband communications and global coverage for ground users. One of the primary challenges for LSN systems with inter-satellite links is the routing strategy calculation and maintenance, due to LSN constellation scale and dynamic network topology feature. In order to seek an efficient routing strategy, a Q-learning-based dynamic distributed Routing scheme for LSNs (QRLSN) is proposed in this paper. To achieve low end-to-end delay and low network traffic overhead load in LSNs, QRLSN adopts a multi-objective optimization method to find the optimal next hop for forwarding data packets. Experimental results demonstrate that the proposed scheme can effectively discover the initial routing strategy and provide long-term Quality of Service (QoS) optimization during the routing maintenance process. In addition, comparison results demonstrate that QRLSN is superior to the virtual-topology-based shortest path routing algorithm.     

Surrogate role of machine learning in motor-drive optimization for more-electric aircraft applications

Motor drives form an essential part of the electric compressors, pumps, braking and actuation systems in the More-Electric Aircraft (MEA). In this paper, the application of Machine Learning (ML) in motor-drive design and optimization process is investigated. The general idea of using ML is to train surrogate models for the optimization. This training process is based on sample data collected from detailed simulation or experiment of motor drives. However, the Surrogate Role (SR) of ML may vary for different applications. This paper first introduces the principles of ML and then proposes two SRs (direct mapping approach and correction approach) of the ML in a motor-drive optimization process. Two different cases are given for the method comparison and validation of ML SRs. The first case is using the sample data from experiments to train the ML surrogate models. For the second case, the joint-simulation data is utilized for a multi-objective motor-drive optimization problem. It is found that both surrogate roles of ML can provide a good mapping model for the cases and in the second case, three feasible design schemes of ML are proposed and validated for the two SRs. Regarding the time consumption in optimizaiton, the proposed ML models can give one motor-drive design point up to 0.044 s while it takes more than 1.5 mins for the used simulation-based models.