Interactivity in map learning: The effect of cognitive load

The hypothesis that active learning is beneficial relative to passive observation was assessed in the context of spatial knowledge derived from maps. Active and passive participants studied a map either while performing a simultaneous spatial tapping task (high cognitive load) or in the absence of this task (low cognitive load). Active participants controlled how the map was learned, with passive participants observing map learning without exercising control. Spatial recall was assessed in two tests, directional judgements and map drawing. Map drawing and directional judgments showed a similar pattern of results, with performance detrimentally affected by a high load for active participants, but not for passive participants. The results indicate that activity and cognitive load interact, suggesting that active learning can be detrimental to spatial learning in cognitively demanding tasks.     

基于梯度增强型Kriging模型的气动反设计方法

基于Kriging模型的代理优化算法目前在气动优化设计中得到了广泛应用。但在高维(设计变量大于30个)气动优化中,计算量过大的问题对其进一步发展产生了严重制约。将翼型和机翼气动反设计问题转化为优化问题,采用Adjoint方法进行快速梯度求解,利用基于梯度增强型Kriging(GEK)模型的代理优化算法分别开展了18、36和108个设计变量的气动反设计。首先,通过采用在设计空间局部建立GEK模型的方法成功地将基于代理优化算法的气动反设计问题的维度拓展到了100维以上。其次,研究了梯度计算精度对基于GEK模型的反设计的影响,发现梯度精度越高,反设计的最终效果越好,同时效率相当。最后,通过不同维度的气动反设计算例,比较了改进拟牛顿法(BFGS)、基于GEK模型和Kriging模型的代理气动反设计方法,结果表明基于GEK模型的代理优化算法的效率大幅度高于基于Kriging模型的代理优化算法,并且维度越高,效率优势越明显;同时,基于GEK模型的代理优化算法在优化效果及分析程序调用次数上相比于BFGS方法也略有优势。     

有源隐身技术研究

讨论了一种新的电子战技术－－有源隐身技术，它可以分成遥地消和人为盲区两种。从理论上研究了有源隐身的基本原理、特点和关键技术，基于电波干扰理论建立的相干有源隐身技术是一种有希望的隐身技术，有源对消技术可使目标的雷达散射面积减缩近２０display status     

基于深度强化学习的高超声速飞行器动态面控制方法

针对高超声速飞行器控制问题,通过将深度强化学习与动态面控制方法相结合,设计了智能姿态控制算法。首先,基于模型先验知识,采用传统的动态面控制方法设计控制器结构。然后,考虑跟踪误差和控制量幅值约束的情况下,采用深度学习算法完成对控制器参数的智能寻优,代替传统设计中的人工调参试错过程。为提升训练效果,在奖励函数中引入了控制量变化率。最后,通过数值仿真验证了所设计控制方法的有效性与鲁棒性。     

涡轮叶片的气动-热-结构多学科设计优化研究

涡轮叶片设计是典型的多学科问题,在保证结果精度的同时必须重视计算效率.通过控制样本的数量和质量,近似模型能够在保证一定精度的前提下,简化多学科优化过程中的数据管理,并提高优化效率.通过分析涡轮转子叶片的气动-热-结构三个学科的设计过程和数据传递关系,充分利用各学科现有的分析工具,建立了涡轮叶片的气动-热-结构多学科优化设计框架.对某涡轮转子叶片分别使用Kriging近似模型和精确分析方法进行优化对比,可以看出两者的结果误差约为1.86%,而效率提高了将近46%,表明采用近似方法的优化结果在工程上是可用的,而且在计算效率更占优势.      

非线性系统在线模糊建模的快速算法

对于复杂、病态、非线性动态系统,基于模糊集合的模糊模型,利用模糊推理规则描述动态系统的特性,是一种有效方法。讨论了利用模糊方法实现非线性系统的建模方法。首先,利用在线模糊竞争学习方法划分输入变量的模糊空间,然后利用卡尔曼滤波算法辨识模糊模型的参数。仿真结果表明了该方法的实用性和有效性。     

Electrostatic space radiation shielding

For the success of NASA’s new vision for space exploration to Moon, Mars and beyond, exposures from the hazards of severe space radiation in deep space long duration missions is ‘a must solve’ problem. The payload penalty demands a very stringent requirement on the design of the spacecrafts for human deep space missions. The exploration beyond low Earth orbit (LEO) to enable routine access of space will require protection from the hazards of the accumulated exposures of space radiation, Galactic Cosmic Rays (GCR) and Solar Particle Events (SPE), and minimizing the production of secondary radiation is a great advantage. There is a need to look to new horizons for newer technologies. The present investigation revisits electrostatic active radiation shielding and explores the feasibility of using the electrostatic shielding in concert with the state-of-the-art materials shielding and protection technologies. The full space radiation environment has been used, for the first time, to explore the feasibility of electrostatic shielding. The goal is to repel enough positive charge ions so that they miss the spacecraft without attracting thermal electrons. Conclusions are drawn for the future directions of space radiation protection.     

Inferring plasma flow velocities from photospheric vector magnetic field observations for the investigation of flare onsets

The amount of emergence and submergence of magnetized plasma and the horizontal motion of the footpoints of flux tubes might be crucial for the dynamics of the solar atmosphere. Although the rate of flux emergence and submergence can be observationally determined near the polarity inversion line (Chae et al., 2004), the same is not true for regions away from the PIL. Also, the horizontal motions cannot be directly measured in the solar photosphere. In this sense, the evolution of the photospheric magnetic field provides valuable information which can be used to estimate photospheric plasma flows since magnetic field and plasma are closely associated (frozen-in-condition). We used three methods to estimate the photospheric plasma motion from magnetic field observations. The methods were applied to photospheric vector magnetic field data of active region NOAA 9077, observed by the Huairou Solar Observing Station (HSOS) of the National Astronomical Observatories of China before and after the ‘Bastille Day’ flare on July 13th and 14th, 2000.     

集成学习在高误码率下AOS协议识别中的应用研究

针对天基网络中高误码率的传输特点,为保证各个异构网络之间数据能够高效可靠的传输,采用空间链路层高级在轨系统协议设计一种基于集成学习的识别方法。该方法采用集成学习模型,学习AOS协议数据,构建基于集成学习的AOS协议识别模型,实现AOS协议的准确识别,并在高误码率情况下进行实验验证。实验结果表明,集成学习模型在AOS协议识别方面具有较好的识别效果,识别运行效率有显著提升,且在较高误码率即10^-1时依然可以保持稳定的识别效果。     

一种基于在轨深度学习的压缩率确定方法

如何对遥感图像中的重要目标进行精确的识别分类,是卫星遥感领域的一个难点和重要的研究方向。深度学习能较好解决识别分类的问题,但其学习模型有大量参数需要确定,会消耗大量计算和存储资源,不利于在轨实现。模型压缩是降低资源需求的有效方法,但会导致分类准确率降低。剪枝是模型压缩的主要方法之一,目前剪枝技术大多研究的是在降低计算量的情况下如何减少准确率损失,如何确定压缩率是有待研究的问题。本文提出了一种通过函数拟合准确率与压缩率关系的方法,可以据此确定相应的压缩率,并对不同的压缩方法进行比较。仿真结果表明:该方法的函数模型可以在不同场景下用较少的点拟合出准确率与压缩率关系曲线,且均方根误差最大为1.09,平均值为0.51,拟合效果较好,可据此针对不同的应用条件与需求确定相应的模型压缩率。