In response to the problems of long training time,slow convergence,and poor performance in some complex situa-tions in deep reinforcement learning in dynamic environments,a combination of multi-step bootstrap,perturbation fluid algorithm(IFDS),and deep deterministic policy gradient algorithm(DDPG)is proposed.Firstly,N-step Bootstrap is added to DDPG to en-dow the model with the ability to combine multiple future time steps.Secondly,the perturbation fluid algorithm is introduced to joint-ly construct a potential field with the velocity information of obstacles in the environment,solving the problem of high-dimensional continuous action space and improving training efficiency.Finally,an environment consisting of a single obstacle and multiple obsta-cles is constructed to simulate and validate the algorithm.The simulation results show that the improved DDPG algorithm has higher training stability and speed compared to traditional DDPG algorithms in the simulation environment,and it can successfully achieve dynamic obstacle avoidance in more complex environments.At the same time,the success rate of training is improved.
The calculation of target distribution area is a prerequisite for anti-ship missile tactical decision-making and target acquisition probability calculation. On the basis of studying the target pursuit model of anti-ship missile, based on the relaxation variable method, the calculation equation for the target distribution is established, sufficient and necessary conditions are proved, the analytical model of target distribution area is obtained. Under the assumption that the target movement direction is arbitrary, and the calculation form under general conditions is given. Through the simulation example, the variation of distribution area under different conditions is compared, which can provide intuitive reference and estimation basis for the decision-making and use of the battle commanders.
In order to improve the professionalism and accuracy of machine translation in the field of notice to air navigation,a two-way machine translation model of notice to air navigation based on transfer learning and back translation is proposed to solve the problem of the parallel corpus of machine translation in the field of air notice.On the basis of training the machine translation of People's Daily,the parameters obtained from the training are transferred to the encoding end and the decoding end of the machine translation model of the partly back-translated announcements to initialize the parameters at both ends,and adjust the parameters at both ends through experiments.The two-way machine translation model of the notice of air travel is obtained.Experiments show that the introduction of the data enhancement strategy of transfer learning and partial back translation improves the robustness of the mod-el and improves the translation quality.The BLEU value of the translation evaluation index of the Chinese-to-English model has in-creased by 2.08%.The BLEU value of the translation evaluation index increased by 3.12%.
ECG is one of the most effective medical tests for heart disease diagnosis, and R-peak detection is the first step in ECG interpretation. For wearable ECG signals, the difficulty of R-peak detection mainly lies in the interference of dynamic strong noise, and the limited hardware computational resources limit the use of some complex algorithms. Therefore, an adaptive threshold R-peak detection algorithm based on Brown's exponential smoothing model is proposed in this paper. The algorithm selects features based on the morphological characteristics and occurrence law of R-peaks, updates the threshold parameters using the Brown exponential smoothing model, optimizes the smoothing coefficients in it using the relative error least squares method, corrects the smoothing ability of the observation error and the response speed to the change of the observation, so that the updated threshold can be more consistent with the R-peak detection. Finally the algorithm achieves 99.6% precision, 99.7% recall and 99.65% F1 score on the self-constructed ECG dataset, and compares with other R-peak detection algorithms to determine the superiority of the proposed algorithm in some performance metrics. It is demonstrated experimentally that the algorithm can adapt well to the strong noise environment and obtain satisfactory R-peak detection accuracy.
无人潜航器协同作战效能评估对于提升潜航器自身作战效能有着重要意义,针对无人潜航器的作战编队,基于DoDAF构建无人潜航器协同作战体系框架,分析作战活动与能力之间的关系,建立协同作战能力指标体系;对于不同类型的指标进行差异化处理,结合ADC模型,考虑综合性指标,减少单项指标的局限性的影响;以对海作战背景下的无人潜航器协同作战为场景,对不同作战方案进行效能评估和计算验证其有效性.
高速热流传感器广泛应用于测量航空发动机涡轮叶片表面热流密度,时间域方面利用时间常数来表征传感器的动态特性.传统的热流传感器动态校准方法,其热流源存在输出功率小,热流不均匀等问题,已然无法满足未来高速热流传感器的动态校准需求.基于高功率光纤输出半导体激光器辐射热流动态校准系统,系统响应时间为μs级,输出热流密度可达50MW/m2.通过Z-t变换测得热流传感器时间常数约为115ms;进行测量不确定度分析,获得扩展不确定度1.804ms,最终获得时间常数最佳估计值τ =(115±1.804)ms,完成薄膜热电堆热流传感器动态测试,并分析传感器动态性能以及综合评价测试平台.
帆船运动模拟器是一种可以在室内进行水上帆船运动的虚拟现实系统,可用于帆船运动员的培训和技能提升,弥补帆船训练的场地限制等问题.模拟器中仿真系统的逼真程度能有效提高帆船技能的学习速度.以OP级帆船为研究对象,从帆船运动的力学原理出发,结合CFD流体仿真对帆翼的空气动力性能进行数值模拟,并以Unity3D为仿真平台,结合3DMax等辅助软件开发了帆船运动仿真系统.与VR Inshore虚拟航行游戏进行对比测试,该仿真系统在一定程度上增加了帆船模拟训练的真实性,并对航速、航线等参数进行了记录,为以后帆船运动员的技能评估提供支持.
小型无人机已经被广泛应用于目标攻击作战任务,但智能化程度较低,仍然非常依赖"人在回路"的远程操作.针对自主攻击任务需求,设计并实现了一套低成本的小型自主攻击无人机系统.首先SoC AI芯片为核心设计了AI处理板卡,完成了目标识别和目标跟踪两种模型的训练和部署,实现了自主攻击中目标检测和跟踪技术;其次设计了目标攻击控制算法;最后搭建了一套小型无人机飞行平台,进行了实际飞行测试.结果表明,整套系统能够对人员、车辆等典型目标进行稳定识别和跟踪锁定,并可实现目标选取和自主攻击.
图像边缘检测是数字图像处理中一个重要的课题.图像边缘检测的目的是找到图像变化强烈的像素点的集合、提取出图像中的重要信息.图像边缘检测在人工智能、图像识别、机器视觉与图像检测等方面有着广泛的应用.论文在Simulink中实现了Roberts、Prewitt、Sobel、Laplace边缘检测算法和高斯滤波算法.给出了使用Simulink进行算法设计与实现的框架.并且在Simulink中对图像边缘检测算法做出调整,提高图像边缘检测的效果.该算法可以利用Simulink与System Generator联合生成HDL IP核,后期在FPGA中直接调用IP核即可.这样可以简化FPGA开发流程,使工作重点放在算法的设计与实现上.
分析了当前电子对抗装备信号场景模拟现状,针对运动场景下雷达目标信号模拟问题,提出了一种运动雷达目标信号模拟方法,根据设置的模拟场景解算三维空间下运动目标模拟信号参数,生成相应的PDW数据,并将其注入R&S SMW200A矢量信号源以合成两路最强的天线模拟信号.由于模拟场景具有可编辑性,该方法可在实验室环境下灵活快捷地模拟各种场景下运动辐射源信号,具有较高的模拟逼真度,同时由于采用PDW数据格式,大大压缩了数据量,能满足高脉冲密度长时间模拟的要求.
针对UUV等水下目标高精度定位跟踪及综合测控要求,开展应答式水声定位研究.提出一种小型应答式水下定位声信标,并基于应答式水声定位系统的软硬件设计开展湖上试验,测试系统的作用距离、测距精度和测向精度,验证了该系统可以正确地测量到信标的距离和方向并对目标进行实时追踪;该系统在湖上的最大作用距离约为1400m,同时具备水声定位跟踪和水声遥测遥控等综合测控功能,具有较高的实用价值和良好的市场应用前景.
过去20年中,复杂网络和电磁环境带来的威胁、人工智能技术的高速发展对指挥和控制(C2)组织体系提出了新的挑战,导致当前的C2组织能力缺乏足够的敏捷性.阐述了C2组织敏捷性的相关概念和理解,开发一系列新的、可实现的概念及解决方案以提高在退化或受限制的网络电磁环境中C2组织的敏捷性.
传统雷达标校方法在效率、适应性、机动性和通用性等方面存在不足,利用经济、快捷、可动态部署的无人机平台标校是雷达标校系统的重要发展方向.论文提出了以系留放飞双模多旋翼无人机、空馈任务载荷与信号源等构建雷达标校系统的方法,分析计算了波瓣图测试与模拟目标测试的射频链路,测试了双模无人机的悬停精度,并通过试验验证了基于双模无人机的波瓣图绘制功能与模拟目标标校功能.采用双模无人机对某舰载雷达进行标定,同时采用传统高塔标校方法进行对照,表明无人机标校系统精度优于雷达测角误差指标的1/5,满足雷达标校需求.
为评估雷达侦察装备的作战能力,结合装备工作原理和改进的模糊层次分析法(FAHP),构建了雷达侦察装备作战能力评估递阶层次结构和评价指标体系,依据各指标相互关系计算确定各指标的权重,根据各指标实际物理意义针对性设计了各指标量化函数,最后通过实例验证了该评估方法的有效性.该评估方法权重计算和指标量化科学合理,具有很强的可操作性,可应用于雷达侦察装备作战试验评估并为试验系统和试验方法设计提供参考.
为提高喷射清理装置的清理效率,基于超音速射流理论,采用半解析方法对喷枪主流道进行了初步设计.在此基础上,采用计算流体力学(Computational Fluid Dynamics,CFD)离散相模型(Discrete phase model,DPM),计算喷枪流道的流动特性,分析了出口直径及颗粒进入位置对喷射效果的影响,据此规律对喷射清理装置喷枪流道进行了优化设计,具有一定的工程指导意义.
控制棒驱动机构(CRDM)转子部件工作在高温高压的冷却剂中,一旦发生故障,其维修或更换的成本高、耗时长、牵连工程大.针对CRDM转子在运行过程中噪声影响过大的问题,提出一种集合经验模态分解(EEMD)结合小波阈值(WTM)的去噪方法,对CRDM转子振动信号进行去噪处理后提取时域特征、频域特征、IMF的能量和样本熵特征等信号作为标签,输入核极限学习机(KELM)分类模型中进行分类识别,并引入了鲸鱼优化算法(WOA)对KELM的重要参数进行自寻优,提高了模型的识别准确率.通过对比实验,证明了论文所提方法在CRDM转子的状态识别上具有比同类方法更高的准确率.
反辐射导弹是压制、摧毁雷达的最有效武器.目前反辐射导弹普遍采用了比幅和相位干涉仪的测向方法,该测向方法易受到诱饵的干扰,使得测向偏离目标.论文将基于四阶累积量的MUSIC方法应用于反辐射导弹的测向,以期实现对时域、频域重叠的雷达和多部诱饵信号的空间分辨,为反辐射导弹抗诱饵干扰提供前提保证,并研究了基于相位模式激励的方法,有效降低了四阶算法的计算量.
针对编队协同作战中存在的任务抢占冲突,构建编队协同作战任务抢占冲突问题模式,并提出单实体状态转移检测算法和任务计划时间一致性调整方法来对任务抢占冲突进行检测与避免.论文的研究结论将为编队作战协同计划的制定提供决策依据.
论文通过分析面阵式激光雷达扫描成像机理,建立飞行目标运动数学模型,采用双切片法对目标模型进行可见点提取,实现面阵式激光雷达对飞行目标的模拟观测扫描.受限于面阵激光雷达分辨率,仅可采集到目标表面部分点云,文中利用凸包算法、灰度图和深度像融合算法对目标点云进行图像增强和滤波处理,以利用缺失点云重构飞行目标轮廓.仿真实验表明,图像增强处理能使导弹弹体和轨迹更加清晰可辨,解决目标飞行速度较快造成的运动模糊和拖尾现象,为后续弹目交会参数解算提供技术支持.
马赛克战是利用自组织网络构建的高度分散、编组灵活、自主协同的杀伤网,使己方拥有更多选择,而让对手陷于复杂和不确定的战场迷雾,在体系对抗中获取不对称优势.论文基于马赛克战概念、特点以及对水面舰艇的现实威胁,分别从态势感知、反隐身、电磁战、辅助决策、装备升级、专项训练及效能评估等方面,给出海军应对马赛克战作战样式的能力建设要点,研究内容可为马赛克战应对策略顶层设计和部队战法训法创新提供参考.
