基于6LoWPAN与MQTT的无线传感网络设计

针对目前6LoWPAN无线传感网络设计与应用存在的可扩展性较差、与IPv4环境不兼容等问题,设计并实现了基于6LoWPAN与消息队列遥控传输(Message Queuing Telemetry Transport, MQTT)的无线传感网络。其中,无线传感网络节点作为MQTT客户端与MQTT服务器通信,边界路由器通过NAT64(Network Address Translation IPv6 to IPv4)实现无线传感网络节点与IPv4网络的通信。将所设计的无线传感网络应用到智能家居场景,评估了无线传感网络和边界路由器的性能,测试了系统功能的实现。结果表明,无线传感网络具有良好的连通性和稳定性,无线传感节点的数据上传与指令接收稳定可靠,边界路由器工作稳定,系统具有较好的性能。     

图像超分辨率卷积神经网络加速算法

为了实现模型的实时和嵌入式运行,提出了一种轻量级的卷积神经网络结构。通过采用较小的滤波器尺寸和引入深度可分离卷积,可大量减少模型参数,提高模型非线性表达能力;在网络末端引入子像素卷积层,直接从原始低分辨率图像学习到高分辨率图像的映射,计算成本为原来的1/k²(k为放大因子)。在Set5数据集上的实验表明,所提模型的速度较经典的图像超分辨率重建算法速度提高了25.8倍,能够在通用GPU上实时运行,峰值信噪比平均提高了0.17 dB,并且参数只有它的35%。     

基于新型非易失内存的远程零拷贝文件系统

为提升物联网与边缘计算应用中前端节点间的数据访问效率,提出了一种新型远程零拷贝文件系统。该文件系统无须借助特殊硬件,可直接基于通用网卡设备实现零拷贝的数据传输框架;充分利用新型非易失内存的随机访问特性,尽可能减少数据缓存和拷贝,提高数据访问的吞吐量。建立缓冲区池,精简并融合传统网络栈和存储栈,从而缩短文件访问路径,减少软件冗余,降低数据访问延迟。最终提供高带宽、低延迟的远程数据访问性能。测试结果表明,远程零拷贝文件系统比传统网络文件系统减少了42.26%～99.19%的读写延迟,细粒度访问下的吞吐量最高可提升1297倍,显著降低了处理器开销和缓存失效次数。     

基于深度森林的网络流量分类方法

随着网络应用的迅猛发展,流量分类在网络资源分配、流量调度和网络安全等诸多研究领域受到广泛关注。现有的机器学习流量分类方法对流量数据特征的选取和分布要求苛刻,导致在实际应用中的复杂流量场景下分类精确度和稳定度难以提高。为了解决样本特征属性的复杂性给分类性能带来的不利影响,引入了基于深度森林的流量分类方法。该算法通过级联森林和多粒度扫描机制,能够在样本数量规模和特征属性选取规模有限的情况下,有效地提高流量整体分类性能。通过网络流量公开数据集Moore对支持向量机、随机森林和深度森林机器学习算法进行训练和测试,结果表明基于深度森林的网络流量分类器的分类准确率能够达到96. 36%,性能优于其他机器学习模型。     

基于多通道卷积神经网络的磁性舰船目标运动参数估计

传统磁性目标运动估计效果依赖于目标的初始状态信息,为克服这一缺陷,建立磁性运动目标三分量投影模型,并据此生成磁性舰船运动目标在运动速度、航向、信噪比等参数变化情况下的10类目标的训练数据集、验证数据集以及测试数据集。设计多通道卷积神经网络,对目标的正横距离和运动速度进行估计,并比较和分析了不同的学习方式和激活函数对网络性能的影响。结果表明:Adam+tanh组合方式的估计性能要优于其他组合方式,而且对磁性目标运动参数的估计效果比较精确,此方法相较于卡尔曼滤波、粒子滤波等估计算法的优越性在于运算复杂度低以及参数估计不需要目标初始状态信息。     

基于强化学习的移动自组织网络分步路由算法*

移动自组织网络是一种无基础设施、由移动通信节点组成的无线网络,具有高度的动态特性。传统的路由协议并不能适应节点移动性带来的频繁拓扑变化,简单的洪泛路由也会因开销过大降低网络的性能。针对如何在移动自组织网络中自适应地进行路由选择,提出了一种基于强化学习的分步路由选择算法。该算法以最小链路总往返时延为目标,基于强化学习进行路由搜寻,在筛选出符合目标需求节点集合的基础上,结合置信度选择路由。在链路变得不可靠时,数据包被广播给筛选出的邻居节点集来提升路由可靠性并降低开销。对提出的算法在分组到达率和路由开销等主要性能指标进行数值仿真分析,仿真结果表明,提出的分步路由算法相比于基于强化学习的智能鲁棒路由,在降低开销的同时,保持着相当的吞吐率。     

Knowledge-oriented modeling for influencing factors of battle damage in military industrial logistics: An integrated method

Modeling influencing factors of battle damage is one of essential works in implementing military industrial logistics simulation to explore battle damage laws knowledge. However, one of key challenges in designing the simulation system could be how to reasonably determine simulation model input and build a bridge to link battle damage model and battle damage laws knowledge. In this paper, we propose a novel knowledge-oriented modeling method for influencing factors of battle damage in military industrial logistics, integrating conceptual analysis, conceptual modeling, quantitative modeling and simulation implementation. We conceptualize influencing factors of battle damage by using the principle of hierarchical decomposition, thus classifying the related battle damage knowledge logically. Then, we construct the conceptual model of influencing factors of battle damage by using Entity-Relationship approach, thus describing their interactions reasonably. Subsequently, we extract the important influencing factors by using social network analysis, thus evaluating their importance quantitatively and further clarifying the elements of simulation. Finally, we develop an agent-based military industry logistics simulation system by taking the modeling results on influencing factors of battle damage as simulation model input, and obtain simulation model output, i.e., new battle damage laws knowledge, thus verifying feasibility and effectiveness of the proposed method. The results show that this method can be used to support human decision-making and action.     

Arbitrary-oriented target detection in large scene sar images

Target detection in the field of synthetic aperture radar (SAR) has attracted considerable attention of researchers in national defense technology worldwide, owing to its unique advantages like high resolution and large scene image acquisition capabilities of SAR. However, due to strong speckle noise and low signal-to-noise ratio, it is difficult to extract representative features of target from SAR images, which greatly inhibits the effectiveness of traditional methods. In order to address the above problems, a framework called contextual rotation region-based convolutional neural network (RCNN) with multilayer fusion is proposed in this paper. Specifically, aimed to enable RCNN to perform target detection in large scene SAR images efficiently, maximum sliding strategy is applied to crop the large scene image into a series of sub-images before RCNN. Instead of using the highest-layer output for proposal generation and target detection, fusion feature maps with high resolution and rich semantic information are constructed by multilayer fusion strategy. Then, we put forwards rotation anchors to predict the minimum circumscribed rectangle of targets to reduce redundant detection region. Furthermore, shadow areas serve as contextual features to provide extraneous information for the detector identify and locate targets accurately. Experimental results on the simulated large scene SAR image dataset show that the proposed method achieves a satisfactory performance in large scene SAR target detection.     

二维矩阵卷积在向量处理器中的设计与实现

为了加快卷积神经网络模型的计算速度,便于大规模神经网络模型在嵌入式微处理器中的实现,以FT-matrix2000向量处理器体系结构为研究背景,通过对多核向量处理器体系结构的分析和对卷积神经网络算法的深入研究,提出将规模较小的卷积核数据置于标量存储体,尺寸较大的卷积矩阵置于向量存储体的数据布局方案。针对矩阵卷积中数据难以复用的问题,提出根据卷积核移动步长的不同动态可配置的混洗模式,通过对所取卷积矩阵元素进行不同的移位操作,进而大幅提高卷积矩阵数据的复用率。针对二维矩阵卷积由于存在数据相关性进而难以多核并行的问题,提出将卷积矩阵多核共享,卷积核矩阵多核独享的多核并行方案。设计了卷积核尺寸不变、卷积矩阵规模变化和卷积矩阵尺寸不变、卷积核规模变化的两种计算方式,并在主流CPU、GPU、TI6678、FT-matrix2000平台进行了性能对比与分析。实验结果表明:FT-matrix2000相比CPU最高可加速238倍,相比TI6678可加速21倍,相比GPU可加速663 805倍。