融入粒子群优化的UPF算法研究及其导航应用

粒子滤波易出现粒子多样性损失,粒子退化等问题,且在初始未知时,需要粒子数较多,收敛较慢。针对上述缺陷,在UKF和PF滤波基础上,将改进了搜索因子的粒子群优化算法融入U粒子滤波算法,提出了一种新型粒子滤波算法,加速了算法的收敛速度。将其应用于地形匹配导航算法中,仿真结果表明该新型算法明显优于UPF及PF算法,更适应于大误差的初始条件下,收敛速度快,具有更高的导航精度和抗噪特性。     

CFO条件下的OFDM子载波调制样式识别方法

提出了一种基于信号幅度分布特征与多次方谱线特征相结合的调制样式识别算法。该算法主要基于正交解调后的正交频分复用子载波信号的幅度分布特征,采用直方图统计的方法实现多进制相移键控和多进制正交幅度调制识别,用多方次谱特征实现多进制相移键控类的调制识别。相比基于经典的高阶累积量的调制识别算法,具有更好的载波频率残留偏差适应能力,在载波频率偏差条件下,提高了调制识别率;相比循环平稳方法,具有更好的信噪比适应能力。仿真实验结果表明了该方法的有效性,相同的识别率下能适应更低的信噪比。     

文本情报信息筛选与聚类的一种处理方法

目前大数据时代情况下的信息有文本、图像、语音和视频等多种形式,而且信息的容量非常大,怎样高效、正确地筛选、分类和处理、利用这些信息,为决策者提供指挥与控制的科学依据显得尤为重要。据此对文本情报信息提出了一种文本聚类的特征选择以及特征变换的方法,利用单词在文本中的出现次数的概率来选择参与聚类的单词,并且对单词出现概率模型定义了特征变换函数,提高了文本信息的筛选、分类和处理的精度,能快速、准确地提取所需要的情报信息提供给指挥与控制的决策者参考、使用。     

Does defence expenditure have a spillover effect on income inequality? A cross-regional analysis in China

China’s defence expenditure has been growing rapidly along with GDP growth during the past two decades. Meanwhile, the income gap has continued to increase. There are conflicting views regarding whether the defence expenditure is capable of reducing income inequality. Therefore, this paper investigates the existence of any spillover effect of defence expenditure on income inequality, with a special focus on the regional differences among 31 provinces and 7 military regions in China. We extend panel cointegration and the impulse response function by using panel data during the period of 1997–2012. The empirical results show that the defence expenditure has an impact on income inequality, and the effect varies over different regions in China. The defence expenditure has a spillover effect on income inequality in the full sample panel and the southeastern panel. An increase in the defence expenditure does not crowd out social welfare spending due to the high level of economic development and government expenditure. On the contrary, in the northern panel, the effect is opposite because of the unbalanced economic development levels within the panel. Beijing as the capital of China, benefits more from the expansion of defence expenditure thus widening the income gap. In addition, the impulse response analyses further confirm a stronger effect of the defence expenditure on income inequality in the northern and the southeastern panels over a short period.     

LEACH—DRT算法中基于神经网络的数据预测算法

为了保证无线传感器网络中数据的完整性,针对基于LEACH路由协议的动态轮时间算法存在的问题,提出一种基于人工神经网络的数据预测算法。该动态轮时间算法中,部分簇因调整后的轮时间不足以完成数据的采集而丢失数据。数据预测算法结合传感器节点数据具有时空相关性的特点,将时空延迟算子引入神经网络模型,并通过建立的神经网络模型对数据进行预测。仿真时采用伯克利英特尔实验室的传感器数据,通过Mafl软件对模型进行测试并分析仿真结果。实验结果表明：该算法对连续多个数据的预测效果理想,预测误差始终保持在较低水平。     

军用电子机箱模态分析及测试

对某军用电子机箱的固有频率从有限元分析和模态试验测试两个方面进行研究,并将理论分析结果和模态试验测试结果进行对比,确定了电子机箱的固有模态.为今后电子机箱的设计提供了参考依据,为仿真分析提供了有效的有限元建模方法.     

Numerical investigation of the shockwave overpressure fields of multi-sources FAE explosions

Shockwaves from fuel-air explosive (FAE) cloud explosions may cause significant casualties. The ground overpressure field is usually used to evaluate the damage range of explosion shockwaves. In this paper, a finite element model of multi-sources FAE explosion is established to simulate the process of multiple shockwaves propagation and interaction. The model is verified with the experimental data of a fourfold-source FAE explosion, with the total fuel mass of 340 kg. Simulation results show that the overpressure fields of multi-sources FAE explosions are different from that of the single-source. In the case of multi-sources, the overpressure fields are influenced significantly by source scattering distance and source number. Subsequently, damage ranges of overpressure under three different levels are calculated. Within a suitable source scattering distance, the damage range of multi-sources situation is greater than that of the single-source, under the same amount of total fuel mass. This research provides a basis for personnel shockwave protection from multi-sources FAE explosion.     

重点目标防空作战中多预警机协同探测空域配置

针对重点目标防空作战中预警机空域配置问题,构建了多预警机阵位选择和架数优化配置模型.通过分析重点目标预警线划分,确定预警机防空作战具体任务,从而构建其阵位选择量化模型,并以此为基础,提出了一种在确保自身安全且充分发挥预警机固有探测威力前提下的多预警机并立航线和串接航线空域配置量化模型.实例运算结果表明,该模型能够对比分...     

多核数字信号处理卷积算法并行优化

针对国防科技大学自主研发的异构多核数字信号处理(digital signal processing, DSP)芯片的特征以及卷积算法自身特点,提出了一种面向多核DSP架构的高性能多核并行卷积实现方案。针对1×1卷积提出了特征图级多核并行方案;针对卷积核大于1的卷积提出了窗口级多核并行优化设计,同时提出了逐元素向量化计算的核内并行优化实现。实验结果表明,所提并行优化方法实现单核计算效率最高能达到64.95%,在带宽受限情况下,多核并行扩展效率可达到48.36%～88.52%,在典型网络ResNet50上的执行性能与E5-2640 CPU相比,获得了5.39倍性能加速。     

基于深度强化学习的无人战车自主行为决策

针对高动态强对抗战场环境下,无人战车面临的自主行为决策问题,分析了未来陆战场无人战车实际作战需求,构建了基于马尔可夫决策过程的自主行为决策模型,提出了一种深度强化学习结合行为树的方法,利用行为树的逻辑规则与先验知识降低强化学习问题的难度,保证收敛性和鲁棒性,同时使行为决策模型具有学习能力.构建典型作战场景,验证深度强化学习结合行为树的无人战车自主行为决策方法的有效性.